diff --git a/Endoscopy/jhu/imerse/endosrt/endosrt/README.md b/Endoscopy/jhu/imerse/endosrt/endosrt/README.md new file mode 100644 index 0000000000000000000000000000000000000000..6132271c13150fd977fced5f0c5327dc952a844f --- /dev/null +++ b/Endoscopy/jhu/imerse/endosrt/endosrt/README.md @@ -0,0 +1,209 @@ + + +# [Soft Robotic Guidewire Navigation] - README + +--- + +## πŸ“‹ At a Glance + +*Teleoperated demonstrations of a 5mm-diameter pneumatic soft robotic guidewire navigating to aneurysms in 3D-printed, rigid, planar phantoms.* +--- + +## πŸ“– Dataset Overview + +*Briefly describe the purpose and content of this dataset. What key skills or scenarios does it demonstrate?* +*This dataset contains 1907 trajectories of a single student demonstrator driving a soft robot's tip point into aneurysm cavities, in addition to 140 trajectories +executed by an ACT-based imitation learning policy. There are 36 geometries used for teleoperation and 6 for the autonomous policy rollout, each with two +aneurysms. Experiments are conducted on a table-top with simulated fluoroscopy as image feedback. It includes successful trials and recovery attempts.* + +| | | +| :--- | :--- | +| **Total Trajectories** | `[2047]` | +| **Total Hours** | `[2.1]` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[x] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [x] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* +- Advancing along vessel paths +- Selecting branches at vascular bifurcations +- Positioning inside aneurysm + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [x] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [x] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[x] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-03-01]` to `[2025-04-30]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [x] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + +*For 602 demonstrations, demonstrations are initialized from a failed robot position, the operator tries to drive it back to the intended path.* + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [ ] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +We used 42 unique phantom geometries. In each of the different geometries, the robot starting position and aneurysm locations were slightly different. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `Custon 3D-printed pneumatic soft robotic guidewire` + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `[Basler a2A2448-75ucBAS, 612x512 @ 25fps]` | +| **Pressure Sensor** | `[Elveflow MPS-V2-L-4]` | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [x] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [ ] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [x] **Other** (Please specify: `[None]`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [x] **Other** (Please specify: `[None]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* + +action: [bend_pos, translate_pos, contrast] +- bend_pos: Absolute position of stepper motor lead screw that drives syringe to induce bending (mL) +- translate_pos: Absolute position of stepper motor lead screw position that drives translation of robot's tube (mm) +- contrast: Binary flag (0/1) indicating whether to initiate a contrast injection + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [ ] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [x] **Other** (Please specify: `[Pressure reading]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +observation.state: [bend_pos, translate_pos, bend_pressure] +- bend_pos: Absolute position of stepper motor lead screw that drives syringe to induce bending (mL) +- translate_pos: Absolute position of stepper motor lead screw position that drives translation of robot's tube (mm) +- bend_pressure: Differential pressure of the robot's internal pneumatic channel (mbar) + +### πŸ“‹ Recommended Additional Representations + +*Even if not your primary action/state representation, we strongly encourage including these standardized formats for maximum compatibility:* + +**Recommended Action Fields:** +- **`action.cartesian_absolute`**: Absolute Cartesian pose with absolute quaternions + ``` + [x, y, z, qx, qy, qz, qw, gripper_angle] + ``` + +**Recommended State Fields:** +- **`observation.state.joint_positions`**: Absolute positions for all articulated joints + ``` + [joint_1, joint_2, ..., joint_n] + ``` + + +--- + +## ⏱️ Data Synchronization Approach + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +*We collect image frames from the Basler camera, pressure readings from the Elveflow, and motor positions from the stepper motors in each iteration of the same software control loop in LabVIEW software. The control loop ran at 25 Hz, and offline checks show skew of Β±1 ms across a 5 minute capture. Thus, the camera, pressure readings, and motor positions are guaranteed to be within a 41 ms window. During export to LeRobot, the timestep's timestamp relative to the beginning of the run is written verbatim into the timestamp attribute. +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Noah Barnes]` | +| **Institution** | `[Johns Hopkins University]` | +| **Contact Email** | `[nbarne18@jhu.edu]` | +| **Citation (BibTeX)** | 
@misc{[endosrt],
  author = {Noah Barnes},
  title = {Soft Robotic Guidewire Navigation},
  year = {2025},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| \ No newline at end of file diff --git a/Endoscopy/ut_austin/arts_lab/colonoscope_lerobot/README.md b/Endoscopy/ut_austin/arts_lab/colonoscope_lerobot/README.md new file mode 100644 index 0000000000000000000000000000000000000000..2d93c37d804159608ffcf79229f140ce1706d310 --- /dev/null +++ b/Endoscopy/ut_austin/arts_lab/colonoscope_lerobot/README.md @@ -0,0 +1,120 @@ +# Colonoscopy Robot Dataset + +πŸ“‹ **At a Glance**: Teleoperated demonstrations of a flexible colonoscope robot navigating through silicone colon phantoms. + +## πŸ“– Dataset Overview + +This dataset contains teleoperated trajectories of the colonoscope robot performing insertion and retraction maneuvers through silicone colon phantoms. Data was collected on an NVIDIA Jetson Orin Nano Super (8GB) via manual teleoperation using an XBox Controller, with NDI Aurora electromagnetic tracking for ground truth tip pose. + +| Metric | Value | +|--------|-------| +| Total Trajectories | 1894 | +| Total Frames | 2095587 | +| Total Hours | ~19.4 | +| Data Type | Table-Top Phantom | +| License | CC BY 4.0 | +| Version | 1.0 | + +## 🎯 Tasks & Domain + +**Domain**: Surgical robotics (Flexible Colonoscopy) + +**Demonstrated Skills**: +- Insertion: Navigating the colonoscope through the colon lumen +- Retraction: Withdrawing the colonoscope while maintaining visualization +- Insertion/Retraction while maintaining colon wall visibility (top/bottom/left/right) +- Common failure modes +- Recovering from occlusions, folds etc + + +## πŸ“Š Multiple Phantoms + +This dataset includes data from multiple phantom models: +- **Global Info**: `meta/info.json` contains mappings (ID ↔ Name) for Phantoms and Sets. +- **Episode Context**: `meta/episodes_context.json` links each episode to its specific Phantom and Set ID. + +## πŸ”¬ Data Collection Details + +| Field | Value | +|-------|-------| +| Collection Method | Human Teleoperation | +| Operator Count | 5 | +| Operator Skill Level | Intermediate (Trained Researcher) | +| Collection Period | From [2025-11-15] to [2026-1-15] +| Recovery Demonstrations | Yes | + +For recovery demonstrations, each episode is initialized from a partially or fully occluded camera position (due to colon folds, closeness to colon walls etc). The operator re-orients the robot such that the lumen is centered again. + +**Diversity Dimensions**: +- [x] Task Execution +- [x] Spatial Layout (varying start positions in phantom) +- [x] Phantom Variation (multiple models) +- [x] Lighting Conditions +- [x] Magnetic Tracker Position + +Each demonstration started and ended at random points in the colon. The tracker position was varied by panning left to right (+/- 20 degrees) as well translating in X and Y (+/- 5 cm). Orientation was fixed along the Z (vertical, gravity aligned) axis. + +## πŸ› οΈ Equipment & Setup + +**Robotic Platform**: Cobra Colonoscope (4-motor tendon-driven flexible endoscope) + +| Sensor Type | Details | +|-------------|---------| +| Primary Camera | Endoscopic Camera, 384x400 @ 30fps via ClearClick Video to USB Capture Device| +| Tracking System | NDI Aurora Electromagnetic Tracker | +| Controller | Xbox Wireless Controller | + +## ⏱️ Data Synchronization + +All sensors are recorded via ROS2 Humble on a single workstation clocked with ROS Time. **Calibration-based time alignment** is applied as a post-processing step using measured sensor lags to ensure all streams represent the same physical state. Specifically, we measure the distinct transport lag of each data stream relative to the camera video stream via sinusoidal fitting. To construct a frame at time $t_{video}$, we sample each sensor stream at a compensated time $t_{sensor}$ to ensure all data corresponds to the same physical instant: + +This accounts for the variable latency between the video feed, the magnetic tracking system, and the motor drivers. + +### A Note on System Dynamics + +The robot is a flexible, tendon-driven colonoscope. Unlike rigid serial manipulators, force transmission from the motors to the tip is not instantaneous. The system exhibits significant mechanical compliance due to: +* **Tendon Slack & Elasticity:** Delay in force propagation. +* **Friction & Tortuosity:** Variable resistance depending on the scope's curvature and interaction with the colon phantom. +* **Backlash:** Hysteresis when changing motion direction. + +Although the static calibration mentioned above has been applied to the system, due to the flexible nature of the system, there is a variable delay betweeen the proximal and distal domain. Specifically, +* `action` (Joystick) and `observation.state` (Motor Encoders) are synchronized. +* `observation.images` (Video) and `observation.ndi_cartesian` (Tip Position) are synchronized. +* The variable delay observed between the Motor Encoders and NDI Tip Position depends on the dynamics of the system. + + +## 🎯 Action & State Space Representation + +### Action Space + +| Field | Shape | Type | Description | +|-------|-------|------|-------------| +| `action` | (4,) | float32 | Control Action: [bend_x, bend_y, insertion, home] | + +### State Space + +| Field | Shape | Type | Description | +|-------|-------|------|-------------| +| `observation.state` | (3,) | int32 | Motor encoder positions | +| `observation.ndi_cartesian_absolute` | (7,) | float32 | Tip pose: x,y,z (meters) + quaternion | +| `observation.ndi_cartesian_relative` | (6,) | float32 | Delta motion: dx,dy,dz + euler | +| `observation.images.endoscope` | (396,383,3) | video | Endoscopic camera view | + + +## πŸ‘₯ Attribution & Contact + +| Field | Value | +|-------|-------| +| Dataset Lead | [Siddhartha Kapuria, Farshid Alambeigi] | +| Department | [Walker Department of Mechanical Engineering] | +| Institution | [The University of Texas at Austin] | +| Contact Email | [skapuria@utexas.edu, farshid.alambeigi@austin.utexas.edu] | + +```bibtex +@misc{cobra_colonoscopy_2026, + author = {[Siddhartha Kapuria, Farshid Alambeigi]}, + title = {Colonoscopy Robot Dataset}, + year = {2026}, + publisher = {Open-H-Embodiment}, +} +``` diff --git a/Surgical/jhu/imerse/nephfat/nephfat/README.md b/Surgical/jhu/imerse/nephfat/nephfat/README.md new file mode 100644 index 0000000000000000000000000000000000000000..3c20f9cfb819cc05661b12cc3cb01fc7f266ac2c --- /dev/null +++ b/Surgical/jhu/imerse/nephfat/nephfat/README.md @@ -0,0 +1,174 @@ + +# NephFat - README + +--- + +## πŸ“‹ At a Glance + +NephFat is a vision-kinematics dataset for perinephric fat dissection in robot-assisted partial nephrectomy, capturing >2,000 trajectories on ex-vivo porcine kidneys using the da Vinci Research Kit-Si (dVRK-Si) and the daVinci Si system. + +--- + +## πŸ“– Dataset Overview + +This proposal introduces a focused, high-quality dataset capturing perinephric fat dissection performed on the da Vinci Si surgical system controlled via the da Vinci Research Kit-Si (dVRK-Si). The task demands precise tissue manipulation, coordinated bimanual tool use, and continuous spatial reasoning. This hierarchical structure supports research in subtask segmentation, skill learning, and long-horizon surgical planning. + +| | | +| :--- | :--- | +| **Total Trajectories** | >2,000 | +| **Total Hours** | | +| **Data Type** | [ ] Clinical
[x] Ex-Vivo
[ ] Table-Top Phantom
[ ] Digital Simulation
[ ] Physical Simulation
[ ] Other (If checked, update "Other") | +| **License** | CC BY 4.0 | +| **Version** | 1.0 (Target Public Release: Mar 2026) | + +--- + +## 🎯 Tasks & Domain + +### Domain + +- [x] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +- Flap Grasp (tissue grasping and retraction) +- Scissors Placing (cutting tool positioning and alignment) +- Cut (controlled tissue dissection) +- Cap Removal (removal of fat overlying the tumor; optional) + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +- [x] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | 3 (Doan Xuan Viet Pham, Dr. Jiawei Ge, Ethan Kilmer) | +| **Operator Skill Level** | [ ] Expert (e.g., Surgeon, Sonographer)
[x] Intermediate (e.g., Trained Researcher)
[x] Novice (e.g., ML Researcher with minimal experience)
[ ] N/A +| **Collection Period** | From 2025-06-01 to 2025-12-31 | + +### Recovery Demonstrations + +- [x] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + +Recovery data was specifically collected for failed scissors placing; in these instances, the 'out-of-distribution' stateβ€”where scissors were misaligned behind or adjacent to the grasped flap rather than correctly positioned for the cutβ€”was deliberately reproduced before recording the corrective recovery trajectory. + +--- + +## πŸ’‘ Diversity Dimensions + +- [ ] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [x] **Other** (Please specify: `Cap Removal`) + +*If you checked any of the above please briefly elaborate below.* + +**Target Object:** The dataset comprises trajectories from $\ge20$ unique tissue samples. Trials were conducted on ex-vivo porcine kidneys prepared with chemically engineered tumor mimics (agarose and cellulose composites). + +**Spatial Layout:** The prepared tumor mimics vary in size, shape and location. + +**Cap Removal:** +Once adequate exposure is achieved, **cap removal is performed optionally**. Cap removal depends on surgeon preference and if existent after fat dissection. Some surgeons retain the fat cap as a grasping handle during subsequent tumor resection. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +- **Robot 1:** da Vinci Si system controlled using the da Vinci Research Kit-Si (dVRK-Si) + +### Sensors & Cameras + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | Stereo endoscopic RGB camera | +| **Room/3rd Person Camera** | - | +| **Force/Torque Sensor** | - | +| **Medical Imager** | - | +| **Other** | Wrist-mounted RGB cameras (left and right arms) | +| **Other** | Robot kinematics and action trajectories | + +--- + +## 🎯 Action & State Space Representation + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [x] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [ ] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [x] **Other** (Please specify: `6D rotation`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [x] **Other** (Please specify: `Hybrid: Position w.r.t Endoscope Camera Tip; Rotation w.r.t End-Effector`) + +**Action Dimensions:** +10-dim action vector for each arm: [dx, dy, dz, r1, r2, r3, r4, r5, r6, jaw] +- dx, dy, dz: Delta position relative to Endoscope Tip Frame (3 dim) +- r1-r6: Delta rotation relative to current End-Effector Frame (6D rotation) +- jaw: Jaw angle (1 dim) + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [x] **Other** (Please specify: `Set Points (_sp), RCM Poses, and Setup Joints (suj)`) + +**State Dimensions:** +*Comprehensive CSV state available (psm1, psm2, ecm, suj). +Key dimensions per arm:* +- **Joints:** 6 dim (`psm*_js[0-5]`) +- **Pose:** 7 dim (`position.x/y/z` + `orientation.x/y/z/w`) +- **Gripper:** 1 dim (`psm*_jaw`) + +--- + +## ⏱️ Data Synchronization Approach + +All sensor streams are time-synchronized, capturing continuous visual observations alongside corresponding robot actions. Quality assurance steps include verification of temporal alignment across modalities and consistency checks for kinematic and image streams. + +--- + +## πŸ‘₯ Attribution & Contact + +| | | +| :--- | :--- | +| **Dataset Lead** | Doan Xuan Viet Pham, Dr. Jiawei Ge, Ethan Kilmer | +| **Institution** | Johns Hopkins University, Technical University of Munich | +| **Contact Email** | viet.x.pham@tum.de, jge9@jhu.edu, ekilmer1@jhu.edu | +| **Citation (BibTeX)** | 
@misc{nephfat_2026,
  author = {Pham, Doan Xuan Viet and Ge, Jiawei and Kilmer, Ethan and Krieger, Axel},
  title = {NephFat: A Vision-Kinematics Dataset for Perinephric Fat Dissection in Robot-Assisted Partial Nephrectomy},
  year = {2026},
  publisher = {Open-H-Embodiment},
}
| \ No newline at end of file diff --git a/Surgical/jhu/lcsr/arcade/cholecystectomy/README.md b/Surgical/jhu/lcsr/arcade/cholecystectomy/README.md new file mode 100644 index 0000000000000000000000000000000000000000..8bf06bba1edfcc7e20f973677b23f01e907d4543 --- /dev/null +++ b/Surgical/jhu/lcsr/arcade/cholecystectomy/README.md @@ -0,0 +1,316 @@ + + +# Cholecystectomy - README + +--- + +## πŸ“‹ At a Glance + +*Teleoperated demonstrations of a da Vinci robot performing Cholecystectomy on a pig liver with galblader* + + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 750 trajectories of novice surgeons using the dVRK to perform Cholecystectomy. It includes successful actions of grasping and dissecting the gallbladder to provide a robust dataset for training imitation learning policies.* + +| | | +| :--- | :--- | +| **Total Trajectories** | `750` | +| **Total Hours** | `75` | +| **Data Type** | `[ ] Clinical` `[X] Ex-Vivo` `[ ] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `2.0` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* + +- Grasping +- Dissection + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `2` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[X] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-11-3]` to `[2025-12-19]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [X] **No** + +**If yes, please briefly describe the recovery process:** + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [X] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [ ] **Target Object** (e.g., different phantom models, suture types) +- [ ] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +The camera position was varied per tissue to simulate different angles of approach. This also leads to different views of the tissue and the tools, which can be used to train policies that are robust to different camera angles. + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +- **Robot 1:** `dVRK (da Vinci Research Kit)` + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Endoscopic Camera, 1920x1080 @ 30fps` | +| **Wrist Cameras** | `CMOS Endoscopy Camera, 516k Pixel (720 x 720) 1mmx1mm Square Camera, 120 Degree FOV, 2.5 m Length, 6 Pin Connector` | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [X] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [X] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [X] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [X] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* + +``` +action.cartesian_psm1: [x, y, z, qx, qy, qz, qw, jaw] +- x, y, z: Absolute position in PSM1 base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- jaw: Jaw/gripper opening angle (radians) +``` + +``` +action.cartesian_psm2: [x, y, z, qx, qy, qz, qw, jaw] +- x, y, z: Absolute position in PSM2 base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- jaw: Jaw/gripper opening angle (radians) +``` + +``` +action.cartesian_ecm: [x, y, z, qx, qy, qz, qw] +- x, y, z: Absolute position in ECM base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion (camera has no gripper) +``` + +``` +action.joint_psm1: [j1, j2, j3, j4, j5, j6] +- j1-j3: Joint positions (radians or meters for prismatic joint) +- j4-j6: Wrist joint positions (radians) +``` + +``` +action.joint_psm2: [j1, j2, j3, j4, j5, j6] +- j1-j3: Joint positions (radians or meters for prismatic joint) +- j4-j6: Wrist joint positions (radians) +``` + +``` +action.joint_ecm: [j1, j2, j3, j4] +- j1-j4: Camera manipulator joint positions (radians or meters) +``` + +``` +Total Action Space (Dual-Arm + ECM): +- Cartesian: 23 dimensions (8 PSM1 + 8 PSM2 + 7 ECM) +- Joint Space: 16 dimensions (6 PSM1 + 6 PSM2 + 4 ECM) +``` + +### State Space Representation + +**State Information Included:** +- [X] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [X] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [X] **Gripper State** (position, force, etc.) +- [X] **Other** (Please specify: `RCM (Remote Center of Motion) poses for PSM1, PSM2, ECM; SUJ (Setup Joints) poses and joint positions for all arms`) + +**State Dimensions:** + +List the state space dimensions and their meanings. + +observation.state: + +PSM1 (Patient Side Manipulator 1): +- psm1_pose: [x, y, z, qx, qy, qz, qw] - End-effector Cartesian pose (meters, quaternion) +- psm1_sp: [x, y, z, qx, qy, qz, qw] - End-effector setpoint/commanded pose +- psm1_jaw: Jaw/gripper opening angle (radians) +- psm1_jaw_sp: Jaw/gripper setpoint angle (radians) +- psm1_rcm_pose: [x, y, z, qx, qy, qz, qw] - Remote Center of Motion pose +- psm1_js: [j1, j2, j3, j4, j5, j6] - Joint positions (radians/meters) +- psm1_set_js: [j1, j2, j3, j4, j5, j6] - Joint setpoints (radians/meters) + +PSM2 (Patient Side Manipulator 2): +- psm2_pose: [x, y, z, qx, qy, qz, qw] - End-effector Cartesian pose +- psm2_sp: [x, y, z, qx, qy, qz, qw] - End-effector setpoint/commanded pose +- psm2_jaw: Jaw/gripper opening angle (radians) +- psm2_jaw_sp: Jaw/gripper setpoint angle (radians) +- psm2_rcm_pose: [x, y, z, qx, qy, qz, qw] - Remote Center of Motion pose +- psm2_js: [j1, j2, j3, j4, j5, j6] - Joint positions (radians/meters) +- psm2_set_js: [j1, j2, j3, j4, j5, j6] - Joint setpoints (radians/meters) + +PSM3 (Patient Side Manipulator 3): +- psm3_js: [j1, j2, j3, j4, j5, j6] - Joint positions (radians/meters) +- psm3_set_js: [j1, j2, j3, j4, j5, j6] - Joint setpoints (radians/meters) + +ECM (Endoscopic Camera Manipulator): +- ecm_pose: [x, y, z, qx, qy, qz, qw] - Camera end-effector pose +- ecm_rcm_pose: [x, y, z, qx, qy, qz, qw] - Remote Center of Motion pose +- ecm_js: [j1, j2, j3, j4] - Joint positions (radians/meters) +- ecm_set_js: [j1, j2, j3, j4] - Joint setpoints (radians/meters) + +SUJ (Setup Joints) - Positioning system for each arm: +- suj1_pose: [x, y, z, qx, qy, qz, qw] - SUJ1 end pose +- suj1_jp: [j1, j2, j3, j4] - SUJ1 joint positions (radians) +- suj2_pose: [x, y, z, qx, qy, qz, qw] - SUJ2 end pose +- suj2_jp: [j1, j2, j3, j4] - SUJ2 joint positions (radians) +- suj3_pose: [x, y, z, qx, qy, qz, qw] - SUJ3 end pose +- suj3_jp: [j1, j2, j3, j4] - SUJ3 joint positions (radians) +- suj_ecm_pose: [x, y, z, qx, qy, qz, qw] - SUJ ECM end pose +- suj_ecm_jp: [j1, j2, j3, j4] - SUJ ECM joint positions (radians) + +Total State Dimension: 148 values + + +### πŸ“‹ Recommended Additional Representations + +*Even if not your primary action/state representation, we strongly encourage including these standardized formats for maximum compatibility:* + +**Recommended Action Fields:** +- **`action.cartesian_absolute_psm1`**: Absolute Cartesian pose for PSM1 + ``` + [x, y, z, qx, qy, qz, qw, jaw_angle] + ``` + +- **`action.cartesian_absolute_psm2`**: Absolute Cartesian pose for PSM2 + ``` + [x, y, z, qx, qy, qz, qw, jaw_angle] + ``` + +- **`action.cartesian_absolute_ecm`**: Absolute Cartesian pose for ECM + ``` + [x, y, z, qx, qy, qz, qw] + ``` + +**Recommended State Fields:** +- **`bservation.state.joint_positions_psm1`**: Absolute positions for PSM1 joints + ``` + [joint_1, joint_2, joint_3, joint_4, joint_5, joint_6] + ``` + +- **`bservation.state.joint_positions_psm2`**: Absolute positions for PSM2 joints + ``` + [joint_1, joint_2, joint_3, joint_4, joint_5, joint_6] + ``` + +- **`bservation.state.joint_positions_ecm`**: Absolute positions for ECM joints + ``` + [joint_1, joint_2, joint_3, joint_4] + ``` +--- + + +Based on the provided scripts, here's the filled-in documentation for your Data Synchronization Approach: + +--- + +## ⏱️ Data Synchronization Approach + +*We capture robot kinematics data and RGB images from multiple camera views (left, right, and two endoscopic cameras), storing timestamps in nanosecond precision directly within image filenames (format: `frame{timestamp_ns}_{camera}.jpg`) and kinematics CSV files. All sensors record timestamps from the same system clock during data collection.* + +**Synchronization Pipeline:** + +1. **Image-to-Kinematics Sync**: For each image timestamp extracted from filenames, we find the nearest kinematics data point in the sorted timestamp array. We check both the floor and ceiling indices and select the closest match by absolute time difference. + +2. **Outlier Filtering**: Frames where the image-to-kinematics time difference exceeds a configurable threshold (default: 30 ms) are marked as outliers and removed from the dataset to ensure temporal alignment quality. + +3. **Multi-Camera Synchronization**: Using the left camera as the temporal reference, we perform binary search to find matching frames across all camera views. A frame is retained only if **all cameras** have a corresponding image within the synchronization tolerance window. This strict enforcement ensures complete multi-view temporal alignment. + +4. **Validation and Export**: The filtering pipeline preserves only fully synchronized frames where both camera alignment and kinematics matching criteria are satisfied. Secondary camera frames are renamed to match the left camera's timestamp, maintaining 1:1 correspondence across all modalities. + +5. **Name and timestamp normalization**: Lastly, we normalize the name of the files to be indexed and the timestamps to be normalized to the start of the episode. This is done by finding the minimum timestamp across all modalities and subtracting it from all timestamps. This ensures that the first frame is always at timestamp 0. + +--- + +--- + +## πŸ‘₯ Attribution & Contact + +| | | +| :--- | :--- | +| **Dataset Lead** | `Jacob M. Delgado LΓ³pez` | +| **Institution** | `Johns Hopkins University` | +| **Contact Email** | `jdelga16@jh.edu` | +| **Citation (BibTeX)** | 
@misc{exvivo_chole_2025,
  author = {Jacob M. Delgado LΓ³pez, Hao Ding, Lalithkumar Seenivasan, Han Zhang, Antony Goldenberg, Juo-Tung Chen, Xinhao Chen, Idris Sunmola, Mathias Unberath},
  title = {Ex-Vivo Porcine Cholecystectomy Subtasks for Multimodal VLA Training},
  year = {2025},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Surgical/jhu/lcsr/miracle/needle_pick_up/README.md b/Surgical/jhu/lcsr/miracle/needle_pick_up/README.md new file mode 100644 index 0000000000000000000000000000000000000000..8ca169df47384dd4771a159beddfef719ca71a5b --- /dev/null +++ b/Surgical/jhu/lcsr/miracle/needle_pick_up/README.md @@ -0,0 +1,209 @@ +# JHU-MIRACLELab/needle-pick-up - README + +--- + +## πŸ“‹ At a Glance + +*Teleoperated demonstrations of a da Vinci robot performing needle pickup on a silicone phantom.* + +--- + +## πŸ“– Dataset Overview + +*This dataset contains TODO trajectory of a trained researcher using the dVRK to perform suture needle pickup.* + +| | | +| :--- | :--- | +| **Total Trajectories** | `1` | +| **Total Hours** | `[Number]` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[V] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [x] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* + +- Needle-pickup + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [x] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `1` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[V] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-12-15]` to `[2025-12-31]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [x] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + +*The operator will keep trying until the needle is successfully picked up.* + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [ ] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [ ] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [x] **Robot Embodiment** (if multiple robots were used) +- [x] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +We placed the needle and tissue phantom differently for each trajectory, and tried different ways of picking up a needle. + +The trajectories contain 50% of a large needle driver as PSM2 and 50% of a cadiere forcep as PSM2. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `dVRK (da Vinci Research Kit)` + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `[e.g., Endoscopic Camera, 1920x1080 @ 30fps]` | +| **Room/3rd Person Camera** | `N/A` | +| **Force/Torque Sensor** | `N/A` | +| **Medical Imager** | `N/A` | +| **Other** | `N/A` | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [x] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [ ] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* + +``` +action: [PSM1 j1, PSM1 j2, PSM1 j3, PSM1 j4, PSM1 j5, PSM1 j6, PSM1 jaw, PSM2 j1, PSM2 j2, PSM2 j3, PSM2 j4, PSM2 j5, PSM2 j6, PSM2 jaw] +- j1-j6: Delta joint angles for the 6 joints of a PSM (radians) +- jaw: Delta jaw angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +``` +observation.state: [ECM j1, ECM j2, ECM j3, ECM j4, ECM x, ECM y, ECM z, ECM qx, ECM qy, ECM qz, ECM qw, PSM1 j1, PSM1 j2, PSM1 j3, PSM1 j4, PSM1 j5, PSM1 j6, PSM1 jaw, +PSM1 x, PSM1 y, PSM1 z, PSM1 qx, PSM1 qy, PSM1 qz, PSM1 qw, +PSM2 j1, PSM2 j2, PSM2 j3, PSM2 j4, PSM2 j5, PSM2 j6, PSM2 jaw, +PSM2 x, PSM2 y, PSM2 z, PSM2 qx, PSM2 qy, PSM2 qz, PSM2 qw] +``` + +### πŸ“‹ Recommended Additional Representations + +*Even if not your primary action/state representation, we strongly encourage including these standardized formats for maximum compatibility:* + +**Recommended Action Fields:** +- **`action.cartesian_absolute`**: Absolute Cartesian pose with absolute quaternions + ``` + [x, y, z, qx, qy, qz, qw, gripper_angle] + ``` + +**Recommended State Fields:** +- **`observation.state.joint_positions`**: Absolute positions for all articulated joints + ``` + [joint_1, joint_2, ..., joint_n] + ``` + + +--- + +## ⏱️ Data Synchronization Approach + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +**Example:** *We collect joint kinematics from our Franka Research 3 and RGB-D frames from Intel RealSense D435 cameras, all running in ROS 2 Galactic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /camera/*/image_raw, and /camera/*/camera_info in a single rosbag2 session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±2 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Chang Liu, Zih-Yun Chiu]` | +| **Institution** | `[Johns Hopkins University]` | +| **Contact Email** | `[cliu250@jh.edu, zchiu@jhu.edu]` | +| **Citation (BibTeX)** | 
@misc{[your_dataset_name_2025],
  author = {[Your Name(s)]},
  title = {[Your Dataset Title]},
  year = {2025},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Surgical/jhu/lcsr/miracle/needle_regrasp/README.md b/Surgical/jhu/lcsr/miracle/needle_regrasp/README.md new file mode 100644 index 0000000000000000000000000000000000000000..28a616d812c0c3d30dfcdf9a75922fe42aef298a --- /dev/null +++ b/Surgical/jhu/lcsr/miracle/needle_regrasp/README.md @@ -0,0 +1,209 @@ +# JHU-MIRACLELab/needle-regrasp - README + +--- + +## πŸ“‹ At a Glance + +*Teleoperated demonstrations of a da Vinci robot regrasping a needle until reaching the optimal configuration.* + +--- + +## πŸ“– Dataset Overview + +*This dataset contains TODO trajectory of a trained researcher using the dVRK to perform suture needle regrasp.* + +| | | +| :--- | :--- | +| **Total Trajectories** | `1` | +| **Total Hours** | `[Number]` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[V] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [x] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* + +- Needle-regrasp + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [x] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `1` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[V] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-12-15]` to `[2025-12-31]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [x] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + +*The operator will keep trying until the needle is successfully grasped at the optimal configuration.* + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [ ] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [ ] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [x] **Robot Embodiment** (if multiple robots were used) +- [x] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +The needle is initially in a random configuration. + +The trajectories contain 50% of a large needle driver as PSM2 and 50% of a cadiere forcep as PSM2. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `dVRK (da Vinci Research Kit)` + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `[e.g., Endoscopic Camera, 1920x1080 @ 30fps]` | +| **Room/3rd Person Camera** | `N/A` | +| **Force/Torque Sensor** | `N/A` | +| **Medical Imager** | `N/A` | +| **Other** | `N/A` | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [x] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [ ] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* + +``` +action: [PSM1 j1, PSM1 j2, PSM1 j3, PSM1 j4, PSM1 j5, PSM1 j6, PSM1 jaw, PSM2 j1, PSM2 j2, PSM2 j3, PSM2 j4, PSM2 j5, PSM2 j6, PSM2 jaw] +- j1-j6: Delta joint angles for the 6 joints of a PSM (radians) +- jaw: Delta jaw angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +``` +observation.state: [ECM j1, ECM j2, ECM j3, ECM j4, ECM x, ECM y, ECM z, ECM qx, ECM qy, ECM qz, ECM qw, PSM1 j1, PSM1 j2, PSM1 j3, PSM1 j4, PSM1 j5, PSM1 j6, PSM1 jaw, +PSM1 x, PSM1 y, PSM1 z, PSM1 qx, PSM1 qy, PSM1 qz, PSM1 qw, +PSM2 j1, PSM2 j2, PSM2 j3, PSM2 j4, PSM2 j5, PSM2 j6, PSM2 jaw, +PSM2 x, PSM2 y, PSM2 z, PSM2 qx, PSM2 qy, PSM2 qz, PSM2 qw] +``` + +### πŸ“‹ Recommended Additional Representations + +*Even if not your primary action/state representation, we strongly encourage including these standardized formats for maximum compatibility:* + +**Recommended Action Fields:** +- **`action.cartesian_absolute`**: Absolute Cartesian pose with absolute quaternions + ``` + [x, y, z, qx, qy, qz, qw, gripper_angle] + ``` + +**Recommended State Fields:** +- **`observation.state.joint_positions`**: Absolute positions for all articulated joints + ``` + [joint_1, joint_2, ..., joint_n] + ``` + + +--- + +## ⏱️ Data Synchronization Approach + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +**Example:** *We collect joint kinematics from our Franka Research 3 and RGB-D frames from Intel RealSense D435 cameras, all running in ROS 2 Galactic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /camera/*/image_raw, and /camera/*/camera_info in a single rosbag2 session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±2 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Chang Liu, Zih-Yun Chiu]` | +| **Institution** | `[Johns Hopkins University]` | +| **Contact Email** | `[cliu250@jh.edu, zchiu@jhu.edu]` | +| **Citation (BibTeX)** | 
@misc{[your_dataset_name_2025],
  author = {[Your Name(s)]},
  title = {[Your Dataset Title]},
  year = {2025},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Surgical/jhu/lcsr/miracle/prepare_to_pierce/README.md b/Surgical/jhu/lcsr/miracle/prepare_to_pierce/README.md new file mode 100644 index 0000000000000000000000000000000000000000..4e70f0ec7a467437cf5ab804020ae0650f472fa2 --- /dev/null +++ b/Surgical/jhu/lcsr/miracle/prepare_to_pierce/README.md @@ -0,0 +1,209 @@ +# JHU-MIRACLELab/prepare-to-pierce - README + +--- + +## πŸ“‹ At a Glance + +*Teleoperated demonstrations of a da Vinci robot grasping the tissue to prepare for piercing.* + +--- + +## πŸ“– Dataset Overview + +*This dataset contains TODO trajectory of a trained researcher using the dVRK to grasp the tissue and prepare for piercing.* + +| | | +| :--- | :--- | +| **Total Trajectories** | `2` | +| **Total Hours** | `[Number]` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[V] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [x] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* + +- Tissue-stabilization + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [x] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `1` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[V] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-12-15]` to `[2025-12-31]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [x] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + +*The operator will keep trying until the tissue is probably stabilized.* + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [ ] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [ ] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [x] **Robot Embodiment** (if multiple robots were used) +- [x] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +The needle is initially in a random configuration. + +The trajectories contain 50% of a large needle driver as PSM2 and 50% of a cadiere forcep as PSM2. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `dVRK (da Vinci Research Kit)` + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `[e.g., Endoscopic Camera, 1920x1080 @ 30fps]` | +| **Room/3rd Person Camera** | `N/A` | +| **Force/Torque Sensor** | `N/A` | +| **Medical Imager** | `N/A` | +| **Other** | `N/A` | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [x] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [ ] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* + +``` +action: [PSM1 j1, PSM1 j2, PSM1 j3, PSM1 j4, PSM1 j5, PSM1 j6, PSM1 jaw, PSM2 j1, PSM2 j2, PSM2 j3, PSM2 j4, PSM2 j5, PSM2 j6, PSM2 jaw] +- j1-j6: Delta joint angles for the 6 joints of a PSM (radians) +- jaw: Delta jaw angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +``` +observation.state: [ECM j1, ECM j2, ECM j3, ECM j4, ECM x, ECM y, ECM z, ECM qx, ECM qy, ECM qz, ECM qw, PSM1 j1, PSM1 j2, PSM1 j3, PSM1 j4, PSM1 j5, PSM1 j6, PSM1 jaw, +PSM1 x, PSM1 y, PSM1 z, PSM1 qx, PSM1 qy, PSM1 qz, PSM1 qw, +PSM2 j1, PSM2 j2, PSM2 j3, PSM2 j4, PSM2 j5, PSM2 j6, PSM2 jaw, +PSM2 x, PSM2 y, PSM2 z, PSM2 qx, PSM2 qy, PSM2 qz, PSM2 qw] +``` + +### πŸ“‹ Recommended Additional Representations + +*Even if not your primary action/state representation, we strongly encourage including these standardized formats for maximum compatibility:* + +**Recommended Action Fields:** +- **`action.cartesian_absolute`**: Absolute Cartesian pose with absolute quaternions + ``` + [x, y, z, qx, qy, qz, qw, gripper_angle] + ``` + +**Recommended State Fields:** +- **`observation.state.joint_positions`**: Absolute positions for all articulated joints + ``` + [joint_1, joint_2, ..., joint_n] + ``` + + +--- + +## ⏱️ Data Synchronization Approach + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +**Example:** *We collect joint kinematics from our Franka Research 3 and RGB-D frames from Intel RealSense D435 cameras, all running in ROS 2 Galactic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /camera/*/image_raw, and /camera/*/camera_info in a single rosbag2 session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±2 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Chang Liu, Zih-Yun Chiu]` | +| **Institution** | `[Johns Hopkins University]` | +| **Contact Email** | `[cliu250@jh.edu, zchiu@jhu.edu]` | +| **Citation (BibTeX)** | 
@misc{[your_dataset_name_2025],
  author = {[Your Name(s)]},
  title = {[Your Dataset Title]},
  year = {2025},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Surgical/jhu/lcsr/smarts/SurgSync-multitask/P1/README.md b/Surgical/jhu/lcsr/smarts/SurgSync-multitask/P1/README.md new file mode 100644 index 0000000000000000000000000000000000000000..41b884fd168a37c713c5675e982b30c693f718cd --- /dev/null +++ b/Surgical/jhu/lcsr/smarts/SurgSync-multitask/P1/README.md @@ -0,0 +1,229 @@ +# SurgSync-multitask P1 + +Canonical SMARTS leaf metadata README. + +- Canonical path: `Surgical/jhu/lcsr/smarts/SurgSync-multitask/P1/` +- Source archive mapping: `online_data_part1.zip`. +- This leaf is one canonical part of the broader JHU SMARTS dataset. + +The broader SMARTS dataset includes two sub-datasets: **offline_recorder** and **online_recorder**. + +--- + +## πŸ“‹ At a Glance + +*Provide a one-sentence summary of your dataset.* + +Teleoperated demonstrations of a da Vinci robot (dVRK-Si) performing multiple canonical tasks on ex-vivo tissue or table-top phantoms; ~50% data collected within the Intuitive abdominal dome model. All data collection uses a modern stereo chip-on-tip endoscope. Both endoscope camera calibration and robot hand-eye calibration will be provided. + +--- + +## File Structure + +For the dataset, it should + +```text +./offline_recorder or online_recorder +β”œβ”€β”€ calibration/ +β”‚ β”œβ”€β”€ case-*... +β”‚ β”‚ β”œβ”€β”€ camera calibration +β”‚ β”‚ β”‚ β”œβ”€β”€ left.yaml +β”‚ β”‚ β”‚ β”œβ”€β”€ right.yaml +β”‚ β”‚ β”‚ └── stereo_calib_params.json +β”‚ β”‚ └── hand_eye_calibration +β”‚ β”‚ β”‚ β”œβ”€β”€ PSM1/2-registration-dVRK.json +β”‚ β”‚ β”‚ └── PSM1/2-registration-open-cv.json +β”œβ”€β”€ data/ +β”‚ └── case-*... +β”œβ”€β”€ videos/ +β”‚ └── case-*... +β”œβ”€β”€ meta/ +β”‚ β”œβ”€β”€ episodes.jsonl +β”‚ β”œβ”€β”€ episodes_stats.jsonl +β”‚ β”œβ”€β”€ tasks.jsonl +β”‚ β”œβ”€β”€ info.json +β”‚ └── README.md +└── total_time.json +``` + +--- + +## πŸ“– Dataset Overview + +*Briefly describe the purpose and content of this dataset. What key skills or scenarios does it demonstrate?* + +This dataset comprises 2500+ teleoperated expert demonstrations of multiple canonical tasks on ex-vivo tissue or table-top phantoms using a surgical robotic system. The tasks include simple interrupted stitch, cold-cut dissection, peg transfer and tissue manipulation. It includes stereo endoscope calibration and robot hand–eye calibration (use dVRK camera registration format), along with high-fidelity, sharp stereo endoscopic video captured via a modern chip-on-tip endoscope. In addition to full task executions, the dataset provides consecutive subtask-level trajectories, enabling analysis of skill composition and procedural structure. Overall, it supports research in imitation learning and skill learning, surgical gesture/subtask recognition, and advanced perception (e.g., tool/tissue interaction understanding and visual tracking) in realistic surgical scenarios + +| | | +| :--- |:-------------------------------------------------------------------------------------------------------------------------------------------------| +| **Total Trajectories** | `1087 ex-vivo (offline) + 1061 ex-vivo (online) + 361 phantom (online)` | +| **Total Hours** | `2.83 ex-vivo (offline) + 3.31 ex-vivo (online) + 0.35 phantom (online)` | +| **Data Type** | `[ ] Clinical` `[X] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +**Note:** The user study experiments have been conducted under HIRB00000701 at Johns Hopkins University. + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* + +The primary skills or procedures demonstrated in this dataset include but not limited to: + +- simple interrupted stitching and its subtasks +- cold cut dissection and its subtasks +- peg transfer and its subtasks +- tissue manipulation and its subtasks +- ... + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Operator Count** | `[13]` | +| **Operator Skill Level** | `[4] Expert (e.g., Surgeon, Sonographer)`
`[5] Intermediate (e.g., Trained Researcher)`
`[4] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-08-01]` to `[2026-01-08]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [X] **No** + +**If yes, please briefly describe the recovery process:** + +**Note:** For your reference, this dataset includes recovery and failure demonstrations, but they are not labeled. + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [X] **Target Object** (e.g., different phantom models, suture types) +- [X] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [X] **Task Execution** (e.g., different techniques for the same task) +- [X] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +The camera (endoscope) can move with dVRK-Si ECM (endoscopy camera manipulator) as needed. The lighting conditions will be changed due to the camera movements. For the stitch, it has both ex-vivo tissue (chicken breast) and silicon phantom cases. The orientation of the wounds could vary from cases to cases. For cold-cut dissection, the target tissue can be different. The options include chicken drumsticks, thin-slide beef/pork, pork belly. The target suture needle could be in various locations. The background can be ex-vivo tissue or the surgical drapes. The detailed subtask approach can vary from cases to cases. For example, the knot-typing could be double throw or single throw. + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `dVRK-Si (the next generation da Vinci Reseach Kit)` + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- |:------------------------------------------------------------------------------------------------------------------------| +| **Primary Camera** | `Endoscopic Camera from Cornerstone Robotics Limited, 1920x1080 @ 60fps, recoreded in 10 FPS` | +| **Room/3rd Person Camera** | `Intel RealSense RGBD camera, only using RGB channel as a mono side-view camera, 1920x1080 @ 30fps, recorded in 10 FPS` | +| **Force/Torque Sensor** | `N/A` | +| **Medical Imager** | `N/A` | +| **Other** | `[Specify]` | + +**Note** The camera calibration files and hand-eye calibration matrices are just for the primary camera. + +--- + +## 🎯 Action & State Space Representation (will update if needed) + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +**Please refer to the subfolder README.md for more details.** + +--- + +## ⏱️ Data Synchronization Approach + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +We use a time-synchronized multi-modal data collection framework from an accepted ICRA 2026 paper: +``` +@inproceedings{zhou2026surgsync, + title={SurgSync: Time-Synchronized Multi-modal Data Collection Framework and Dataset for Surgical Robotics}, + author={Zhou, Haoying and ... and Kazanzides, Peter}, + booktitle={IEEE Intl. Conf. on Robotics and Automation (ICRA)}, + year={2026} +} +``` +We have two desktop: one for dVRK running and one for all the other pipelines. They are connected using ROS network and its NTP. + +We have two modes when data collection, and the performance is highly dependent on the hardware. + +**Online(-matching) Recorder**: (not uploaded yet) + +The design enforces strict time synchronization using multi-threading: only samples that fall within a user-defined time tolerance are admitted. This yields slightly uneven inter-sample intervals (irregular dt), +but it retains the natural continuity of smooth teleoperation segments and avoids label/feature drift. In our study, a time tolerance of 10 ms is selected to ensure both high time +alignment tightness and consecutive recorder output. + +As a result, the recorder time latency is 6.36 (+/- 4.72) ms with a median of 5.58ms. The recording frequency is 4.04 (+\- 1.69) Hz. We will assemble everything assume 10 FPS. + +**Offline(-matching) Recorder**: (already fully uploaded) + +Our offline-matching approach decouples recording from time alignments to maximize +the recording system efficiency. Therefore, this recorder produces synchronized datasets in two stages: (i) a lightweight +recorder logs camera streams to videos (pre-synced) and raw kinematic streams to binary files to disk with minimal processing; +(ii) an offline post-processing pipeline reconstructs a fixed-rate frame sequence and, for each frame, gathers the five +closest samples using nearest-timestamp lookup for subsequent interpolation. Compared to the online-matching recorder (which +pairs visual and kinematic data in real time), this two-stage design avoids tolerance-based dropping of data during capture +yielding a higher throughput and uniform intervals between synchronized data packets at the cost of requiring more storage +and substantial time for post-collection time-matching and interpolation. + +As a result, the recorder time latency is 1.35 (+/- 0.81) ms with a median of 1.33ms. And the recording frequency is a solid 10.00 FPS. + +**Note:** *Both recorders may have few outliners due to our suboptimal workstation hardware.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Dataset Lead** | `[Haoying Zhou (student lead), Peter Kazanzides (faculty PI)]` | +| **Institution** | `[Johns Hopkins University, Worcester Polytechnic Institute (Haoying Zhou)]` | +| **Contact Email** | `[hzhou6@wpi.edu, hzhou62@jh.edu, pkaz@jhu.edu]` | +| **Citation (BibTeX)** | 
@misc{[SurgSyncExt],
  author = {[Haoying Zhou, Peter Kazanzides, Chang Liu, Junlin Wu, Yimeng Wu, Hongjun Wu]},
  title = {[SurgSyncExt: Time-Synchronized Multi-Modal Dataset for Surgical Robotics]},
  year = {2025},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Surgical/jhu/lcsr/smarts/SurgSync-multitask/P2/README.md b/Surgical/jhu/lcsr/smarts/SurgSync-multitask/P2/README.md new file mode 100644 index 0000000000000000000000000000000000000000..2a501c9cac58f09ff4b600f277ffa956f65616bd --- /dev/null +++ b/Surgical/jhu/lcsr/smarts/SurgSync-multitask/P2/README.md @@ -0,0 +1,229 @@ +# SurgSync-multitask P2 + +Canonical SMARTS leaf metadata README. + +- Canonical path: `Surgical/jhu/lcsr/smarts/SurgSync-multitask/P2/` +- Source archive mapping: `online_data_part2.zip`. +- This leaf is one canonical part of the broader JHU SMARTS dataset. + +The broader SMARTS dataset includes two sub-datasets: **offline_recorder** and **online_recorder**. + +--- + +## πŸ“‹ At a Glance + +*Provide a one-sentence summary of your dataset.* + +Teleoperated demonstrations of a da Vinci robot (dVRK-Si) performing multiple canonical tasks on ex-vivo tissue or table-top phantoms; ~50% data collected within the Intuitive abdominal dome model. All data collection uses a modern stereo chip-on-tip endoscope. Both endoscope camera calibration and robot hand-eye calibration will be provided. + +--- + +## File Structure + +For the dataset, it should + +```text +./offline_recorder or online_recorder +β”œβ”€β”€ calibration/ +β”‚ β”œβ”€β”€ case-*... +β”‚ β”‚ β”œβ”€β”€ camera calibration +β”‚ β”‚ β”‚ β”œβ”€β”€ left.yaml +β”‚ β”‚ β”‚ β”œβ”€β”€ right.yaml +β”‚ β”‚ β”‚ └── stereo_calib_params.json +β”‚ β”‚ └── hand_eye_calibration +β”‚ β”‚ β”‚ β”œβ”€β”€ PSM1/2-registration-dVRK.json +β”‚ β”‚ β”‚ └── PSM1/2-registration-open-cv.json +β”œβ”€β”€ data/ +β”‚ └── case-*... +β”œβ”€β”€ videos/ +β”‚ └── case-*... +β”œβ”€β”€ meta/ +β”‚ β”œβ”€β”€ episodes.jsonl +β”‚ β”œβ”€β”€ episodes_stats.jsonl +β”‚ β”œβ”€β”€ tasks.jsonl +β”‚ β”œβ”€β”€ info.json +β”‚ └── README.md +└── total_time.json +``` + +--- + +## πŸ“– Dataset Overview + +*Briefly describe the purpose and content of this dataset. What key skills or scenarios does it demonstrate?* + +This dataset comprises 2500+ teleoperated expert demonstrations of multiple canonical tasks on ex-vivo tissue or table-top phantoms using a surgical robotic system. The tasks include simple interrupted stitch, cold-cut dissection, peg transfer and tissue manipulation. It includes stereo endoscope calibration and robot hand–eye calibration (use dVRK camera registration format), along with high-fidelity, sharp stereo endoscopic video captured via a modern chip-on-tip endoscope. In addition to full task executions, the dataset provides consecutive subtask-level trajectories, enabling analysis of skill composition and procedural structure. Overall, it supports research in imitation learning and skill learning, surgical gesture/subtask recognition, and advanced perception (e.g., tool/tissue interaction understanding and visual tracking) in realistic surgical scenarios + +| | | +| :--- |:-------------------------------------------------------------------------------------------------------------------------------------------------| +| **Total Trajectories** | `1087 ex-vivo (offline) + 1061 ex-vivo (online) + 361 phantom (online)` | +| **Total Hours** | `2.83 ex-vivo (offline) + 3.31 ex-vivo (online) + 0.35 phantom (online)` | +| **Data Type** | `[ ] Clinical` `[X] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +**Note:** The user study experiments have been conducted under HIRB00000701 at Johns Hopkins University. + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* + +The primary skills or procedures demonstrated in this dataset include but not limited to: + +- simple interrupted stitching and its subtasks +- cold cut dissection and its subtasks +- peg transfer and its subtasks +- tissue manipulation and its subtasks +- ... + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Operator Count** | `[13]` | +| **Operator Skill Level** | `[4] Expert (e.g., Surgeon, Sonographer)`
`[5] Intermediate (e.g., Trained Researcher)`
`[4] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-08-01]` to `[2026-01-08]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [X] **No** + +**If yes, please briefly describe the recovery process:** + +**Note:** For your reference, this dataset includes recovery and failure demonstrations, but they are not labeled. + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [X] **Target Object** (e.g., different phantom models, suture types) +- [X] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [X] **Task Execution** (e.g., different techniques for the same task) +- [X] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +The camera (endoscope) can move with dVRK-Si ECM (endoscopy camera manipulator) as needed. The lighting conditions will be changed due to the camera movements. For the stitch, it has both ex-vivo tissue (chicken breast) and silicon phantom cases. The orientation of the wounds could vary from cases to cases. For cold-cut dissection, the target tissue can be different. The options include chicken drumsticks, thin-slide beef/pork, pork belly. The target suture needle could be in various locations. The background can be ex-vivo tissue or the surgical drapes. The detailed subtask approach can vary from cases to cases. For example, the knot-typing could be double throw or single throw. + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `dVRK-Si (the next generation da Vinci Reseach Kit)` + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- |:------------------------------------------------------------------------------------------------------------------------| +| **Primary Camera** | `Endoscopic Camera from Cornerstone Robotics Limited, 1920x1080 @ 60fps, recoreded in 10 FPS` | +| **Room/3rd Person Camera** | `Intel RealSense RGBD camera, only using RGB channel as a mono side-view camera, 1920x1080 @ 30fps, recorded in 10 FPS` | +| **Force/Torque Sensor** | `N/A` | +| **Medical Imager** | `N/A` | +| **Other** | `[Specify]` | + +**Note** The camera calibration files and hand-eye calibration matrices are just for the primary camera. + +--- + +## 🎯 Action & State Space Representation (will update if needed) + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +**Please refer to the subfolder README.md for more details.** + +--- + +## ⏱️ Data Synchronization Approach + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +We use a time-synchronized multi-modal data collection framework from an accepted ICRA 2026 paper: +``` +@inproceedings{zhou2026surgsync, + title={SurgSync: Time-Synchronized Multi-modal Data Collection Framework and Dataset for Surgical Robotics}, + author={Zhou, Haoying and ... and Kazanzides, Peter}, + booktitle={IEEE Intl. Conf. on Robotics and Automation (ICRA)}, + year={2026} +} +``` +We have two desktop: one for dVRK running and one for all the other pipelines. They are connected using ROS network and its NTP. + +We have two modes when data collection, and the performance is highly dependent on the hardware. + +**Online(-matching) Recorder**: (not uploaded yet) + +The design enforces strict time synchronization using multi-threading: only samples that fall within a user-defined time tolerance are admitted. This yields slightly uneven inter-sample intervals (irregular dt), +but it retains the natural continuity of smooth teleoperation segments and avoids label/feature drift. In our study, a time tolerance of 10 ms is selected to ensure both high time +alignment tightness and consecutive recorder output. + +As a result, the recorder time latency is 6.36 (+/- 4.72) ms with a median of 5.58ms. The recording frequency is 4.04 (+\- 1.69) Hz. We will assemble everything assume 10 FPS. + +**Offline(-matching) Recorder**: (already fully uploaded) + +Our offline-matching approach decouples recording from time alignments to maximize +the recording system efficiency. Therefore, this recorder produces synchronized datasets in two stages: (i) a lightweight +recorder logs camera streams to videos (pre-synced) and raw kinematic streams to binary files to disk with minimal processing; +(ii) an offline post-processing pipeline reconstructs a fixed-rate frame sequence and, for each frame, gathers the five +closest samples using nearest-timestamp lookup for subsequent interpolation. Compared to the online-matching recorder (which +pairs visual and kinematic data in real time), this two-stage design avoids tolerance-based dropping of data during capture +yielding a higher throughput and uniform intervals between synchronized data packets at the cost of requiring more storage +and substantial time for post-collection time-matching and interpolation. + +As a result, the recorder time latency is 1.35 (+/- 0.81) ms with a median of 1.33ms. And the recording frequency is a solid 10.00 FPS. + +**Note:** *Both recorders may have few outliners due to our suboptimal workstation hardware.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Dataset Lead** | `[Haoying Zhou (student lead), Peter Kazanzides (faculty PI)]` | +| **Institution** | `[Johns Hopkins University, Worcester Polytechnic Institute (Haoying Zhou)]` | +| **Contact Email** | `[hzhou6@wpi.edu, hzhou62@jh.edu, pkaz@jhu.edu]` | +| **Citation (BibTeX)** | 
@misc{[SurgSyncExt],
  author = {[Haoying Zhou, Peter Kazanzides, Chang Liu, Junlin Wu, Yimeng Wu, Hongjun Wu]},
  title = {[SurgSyncExt: Time-Synchronized Multi-Modal Dataset for Surgical Robotics]},
  year = {2025},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Surgical/jhu/lcsr/smarts/SurgSync-multitask/P3/README.md b/Surgical/jhu/lcsr/smarts/SurgSync-multitask/P3/README.md new file mode 100644 index 0000000000000000000000000000000000000000..fafbb474ad304463e4cc4fd7ce4b53ea9ed9d46d --- /dev/null +++ b/Surgical/jhu/lcsr/smarts/SurgSync-multitask/P3/README.md @@ -0,0 +1,229 @@ +# SurgSync-multitask P3 + +Canonical SMARTS leaf metadata README. + +- Canonical path: `Surgical/jhu/lcsr/smarts/SurgSync-multitask/P3/` +- Source archive mapping: `online_data_part3.zip`. +- This leaf is one canonical part of the broader JHU SMARTS dataset. + +The broader SMARTS dataset includes two sub-datasets: **offline_recorder** and **online_recorder**. + +--- + +## πŸ“‹ At a Glance + +*Provide a one-sentence summary of your dataset.* + +Teleoperated demonstrations of a da Vinci robot (dVRK-Si) performing multiple canonical tasks on ex-vivo tissue or table-top phantoms; ~50% data collected within the Intuitive abdominal dome model. All data collection uses a modern stereo chip-on-tip endoscope. Both endoscope camera calibration and robot hand-eye calibration will be provided. + +--- + +## File Structure + +For the dataset, it should + +```text +./offline_recorder or online_recorder +β”œβ”€β”€ calibration/ +β”‚ β”œβ”€β”€ case-*... +β”‚ β”‚ β”œβ”€β”€ camera calibration +β”‚ β”‚ β”‚ β”œβ”€β”€ left.yaml +β”‚ β”‚ β”‚ β”œβ”€β”€ right.yaml +β”‚ β”‚ β”‚ └── stereo_calib_params.json +β”‚ β”‚ └── hand_eye_calibration +β”‚ β”‚ β”‚ β”œβ”€β”€ PSM1/2-registration-dVRK.json +β”‚ β”‚ β”‚ └── PSM1/2-registration-open-cv.json +β”œβ”€β”€ data/ +β”‚ └── case-*... +β”œβ”€β”€ videos/ +β”‚ └── case-*... +β”œβ”€β”€ meta/ +β”‚ β”œβ”€β”€ episodes.jsonl +β”‚ β”œβ”€β”€ episodes_stats.jsonl +β”‚ β”œβ”€β”€ tasks.jsonl +β”‚ β”œβ”€β”€ info.json +β”‚ └── README.md +└── total_time.json +``` + +--- + +## πŸ“– Dataset Overview + +*Briefly describe the purpose and content of this dataset. What key skills or scenarios does it demonstrate?* + +This dataset comprises 2500+ teleoperated expert demonstrations of multiple canonical tasks on ex-vivo tissue or table-top phantoms using a surgical robotic system. The tasks include simple interrupted stitch, cold-cut dissection, peg transfer and tissue manipulation. It includes stereo endoscope calibration and robot hand–eye calibration (use dVRK camera registration format), along with high-fidelity, sharp stereo endoscopic video captured via a modern chip-on-tip endoscope. In addition to full task executions, the dataset provides consecutive subtask-level trajectories, enabling analysis of skill composition and procedural structure. Overall, it supports research in imitation learning and skill learning, surgical gesture/subtask recognition, and advanced perception (e.g., tool/tissue interaction understanding and visual tracking) in realistic surgical scenarios + +| | | +| :--- |:-------------------------------------------------------------------------------------------------------------------------------------------------| +| **Total Trajectories** | `1087 ex-vivo (offline) + 1061 ex-vivo (online) + 361 phantom (online)` | +| **Total Hours** | `2.83 ex-vivo (offline) + 3.31 ex-vivo (online) + 0.35 phantom (online)` | +| **Data Type** | `[ ] Clinical` `[X] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +**Note:** The user study experiments have been conducted under HIRB00000701 at Johns Hopkins University. + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* + +The primary skills or procedures demonstrated in this dataset include but not limited to: + +- simple interrupted stitching and its subtasks +- cold cut dissection and its subtasks +- peg transfer and its subtasks +- tissue manipulation and its subtasks +- ... + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Operator Count** | `[13]` | +| **Operator Skill Level** | `[4] Expert (e.g., Surgeon, Sonographer)`
`[5] Intermediate (e.g., Trained Researcher)`
`[4] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-08-01]` to `[2026-01-08]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [X] **No** + +**If yes, please briefly describe the recovery process:** + +**Note:** For your reference, this dataset includes recovery and failure demonstrations, but they are not labeled. + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [X] **Target Object** (e.g., different phantom models, suture types) +- [X] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [X] **Task Execution** (e.g., different techniques for the same task) +- [X] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +The camera (endoscope) can move with dVRK-Si ECM (endoscopy camera manipulator) as needed. The lighting conditions will be changed due to the camera movements. For the stitch, it has both ex-vivo tissue (chicken breast) and silicon phantom cases. The orientation of the wounds could vary from cases to cases. For cold-cut dissection, the target tissue can be different. The options include chicken drumsticks, thin-slide beef/pork, pork belly. The target suture needle could be in various locations. The background can be ex-vivo tissue or the surgical drapes. The detailed subtask approach can vary from cases to cases. For example, the knot-typing could be double throw or single throw. + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `dVRK-Si (the next generation da Vinci Reseach Kit)` + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- |:------------------------------------------------------------------------------------------------------------------------| +| **Primary Camera** | `Endoscopic Camera from Cornerstone Robotics Limited, 1920x1080 @ 60fps, recoreded in 10 FPS` | +| **Room/3rd Person Camera** | `Intel RealSense RGBD camera, only using RGB channel as a mono side-view camera, 1920x1080 @ 30fps, recorded in 10 FPS` | +| **Force/Torque Sensor** | `N/A` | +| **Medical Imager** | `N/A` | +| **Other** | `[Specify]` | + +**Note** The camera calibration files and hand-eye calibration matrices are just for the primary camera. + +--- + +## 🎯 Action & State Space Representation (will update if needed) + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +**Please refer to the subfolder README.md for more details.** + +--- + +## ⏱️ Data Synchronization Approach + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +We use a time-synchronized multi-modal data collection framework from an accepted ICRA 2026 paper: +``` +@inproceedings{zhou2026surgsync, + title={SurgSync: Time-Synchronized Multi-modal Data Collection Framework and Dataset for Surgical Robotics}, + author={Zhou, Haoying and ... and Kazanzides, Peter}, + booktitle={IEEE Intl. Conf. on Robotics and Automation (ICRA)}, + year={2026} +} +``` +We have two desktop: one for dVRK running and one for all the other pipelines. They are connected using ROS network and its NTP. + +We have two modes when data collection, and the performance is highly dependent on the hardware. + +**Online(-matching) Recorder**: (not uploaded yet) + +The design enforces strict time synchronization using multi-threading: only samples that fall within a user-defined time tolerance are admitted. This yields slightly uneven inter-sample intervals (irregular dt), +but it retains the natural continuity of smooth teleoperation segments and avoids label/feature drift. In our study, a time tolerance of 10 ms is selected to ensure both high time +alignment tightness and consecutive recorder output. + +As a result, the recorder time latency is 6.36 (+/- 4.72) ms with a median of 5.58ms. The recording frequency is 4.04 (+\- 1.69) Hz. We will assemble everything assume 10 FPS. + +**Offline(-matching) Recorder**: (already fully uploaded) + +Our offline-matching approach decouples recording from time alignments to maximize +the recording system efficiency. Therefore, this recorder produces synchronized datasets in two stages: (i) a lightweight +recorder logs camera streams to videos (pre-synced) and raw kinematic streams to binary files to disk with minimal processing; +(ii) an offline post-processing pipeline reconstructs a fixed-rate frame sequence and, for each frame, gathers the five +closest samples using nearest-timestamp lookup for subsequent interpolation. Compared to the online-matching recorder (which +pairs visual and kinematic data in real time), this two-stage design avoids tolerance-based dropping of data during capture +yielding a higher throughput and uniform intervals between synchronized data packets at the cost of requiring more storage +and substantial time for post-collection time-matching and interpolation. + +As a result, the recorder time latency is 1.35 (+/- 0.81) ms with a median of 1.33ms. And the recording frequency is a solid 10.00 FPS. + +**Note:** *Both recorders may have few outliners due to our suboptimal workstation hardware.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Dataset Lead** | `[Haoying Zhou (student lead), Peter Kazanzides (faculty PI)]` | +| **Institution** | `[Johns Hopkins University, Worcester Polytechnic Institute (Haoying Zhou)]` | +| **Contact Email** | `[hzhou6@wpi.edu, hzhou62@jh.edu, pkaz@jhu.edu]` | +| **Citation (BibTeX)** | 
@misc{[SurgSyncExt],
  author = {[Haoying Zhou, Peter Kazanzides, Chang Liu, Junlin Wu, Yimeng Wu, Hongjun Wu]},
  title = {[SurgSyncExt: Time-Synchronized Multi-Modal Dataset for Surgical Robotics]},
  year = {2025},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Surgical/jhu/lcsr/smarts/SurgSync-multitask/P4/README.md b/Surgical/jhu/lcsr/smarts/SurgSync-multitask/P4/README.md new file mode 100644 index 0000000000000000000000000000000000000000..ba38d73e3372f01777f94cb1a487a1a92d9cca39 --- /dev/null +++ b/Surgical/jhu/lcsr/smarts/SurgSync-multitask/P4/README.md @@ -0,0 +1,229 @@ +# SurgSync-multitask P4 + +Canonical SMARTS leaf metadata README. + +- Canonical path: `Surgical/jhu/lcsr/smarts/SurgSync-multitask/P4/` +- Source archive mapping: `online_data_part4.zip`. +- This leaf is one canonical part of the broader JHU SMARTS dataset. + +The broader SMARTS dataset includes two sub-datasets: **offline_recorder** and **online_recorder**. + +--- + +## πŸ“‹ At a Glance + +*Provide a one-sentence summary of your dataset.* + +Teleoperated demonstrations of a da Vinci robot (dVRK-Si) performing multiple canonical tasks on ex-vivo tissue or table-top phantoms; ~50% data collected within the Intuitive abdominal dome model. All data collection uses a modern stereo chip-on-tip endoscope. Both endoscope camera calibration and robot hand-eye calibration will be provided. + +--- + +## File Structure + +For the dataset, it should + +```text +./offline_recorder or online_recorder +β”œβ”€β”€ calibration/ +β”‚ β”œβ”€β”€ case-*... +β”‚ β”‚ β”œβ”€β”€ camera calibration +β”‚ β”‚ β”‚ β”œβ”€β”€ left.yaml +β”‚ β”‚ β”‚ β”œβ”€β”€ right.yaml +β”‚ β”‚ β”‚ └── stereo_calib_params.json +β”‚ β”‚ └── hand_eye_calibration +β”‚ β”‚ β”‚ β”œβ”€β”€ PSM1/2-registration-dVRK.json +β”‚ β”‚ β”‚ └── PSM1/2-registration-open-cv.json +β”œβ”€β”€ data/ +β”‚ └── case-*... +β”œβ”€β”€ videos/ +β”‚ └── case-*... +β”œβ”€β”€ meta/ +β”‚ β”œβ”€β”€ episodes.jsonl +β”‚ β”œβ”€β”€ episodes_stats.jsonl +β”‚ β”œβ”€β”€ tasks.jsonl +β”‚ β”œβ”€β”€ info.json +β”‚ └── README.md +└── total_time.json +``` + +--- + +## πŸ“– Dataset Overview + +*Briefly describe the purpose and content of this dataset. What key skills or scenarios does it demonstrate?* + +This dataset comprises 2500+ teleoperated expert demonstrations of multiple canonical tasks on ex-vivo tissue or table-top phantoms using a surgical robotic system. The tasks include simple interrupted stitch, cold-cut dissection, peg transfer and tissue manipulation. It includes stereo endoscope calibration and robot hand–eye calibration (use dVRK camera registration format), along with high-fidelity, sharp stereo endoscopic video captured via a modern chip-on-tip endoscope. In addition to full task executions, the dataset provides consecutive subtask-level trajectories, enabling analysis of skill composition and procedural structure. Overall, it supports research in imitation learning and skill learning, surgical gesture/subtask recognition, and advanced perception (e.g., tool/tissue interaction understanding and visual tracking) in realistic surgical scenarios + +| | | +| :--- |:-------------------------------------------------------------------------------------------------------------------------------------------------| +| **Total Trajectories** | `1087 ex-vivo (offline) + 1061 ex-vivo (online) + 361 phantom (online)` | +| **Total Hours** | `2.83 ex-vivo (offline) + 3.31 ex-vivo (online) + 0.35 phantom (online)` | +| **Data Type** | `[ ] Clinical` `[X] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +**Note:** The user study experiments have been conducted under HIRB00000701 at Johns Hopkins University. + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* + +The primary skills or procedures demonstrated in this dataset include but not limited to: + +- simple interrupted stitching and its subtasks +- cold cut dissection and its subtasks +- peg transfer and its subtasks +- tissue manipulation and its subtasks +- ... + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Operator Count** | `[13]` | +| **Operator Skill Level** | `[4] Expert (e.g., Surgeon, Sonographer)`
`[5] Intermediate (e.g., Trained Researcher)`
`[4] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-08-01]` to `[2026-01-08]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [X] **No** + +**If yes, please briefly describe the recovery process:** + +**Note:** For your reference, this dataset includes recovery and failure demonstrations, but they are not labeled. + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [X] **Target Object** (e.g., different phantom models, suture types) +- [X] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [X] **Task Execution** (e.g., different techniques for the same task) +- [X] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +The camera (endoscope) can move with dVRK-Si ECM (endoscopy camera manipulator) as needed. The lighting conditions will be changed due to the camera movements. For the stitch, it has both ex-vivo tissue (chicken breast) and silicon phantom cases. The orientation of the wounds could vary from cases to cases. For cold-cut dissection, the target tissue can be different. The options include chicken drumsticks, thin-slide beef/pork, pork belly. The target suture needle could be in various locations. The background can be ex-vivo tissue or the surgical drapes. The detailed subtask approach can vary from cases to cases. For example, the knot-typing could be double throw or single throw. + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `dVRK-Si (the next generation da Vinci Reseach Kit)` + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- |:------------------------------------------------------------------------------------------------------------------------| +| **Primary Camera** | `Endoscopic Camera from Cornerstone Robotics Limited, 1920x1080 @ 60fps, recoreded in 10 FPS` | +| **Room/3rd Person Camera** | `Intel RealSense RGBD camera, only using RGB channel as a mono side-view camera, 1920x1080 @ 30fps, recorded in 10 FPS` | +| **Force/Torque Sensor** | `N/A` | +| **Medical Imager** | `N/A` | +| **Other** | `[Specify]` | + +**Note** The camera calibration files and hand-eye calibration matrices are just for the primary camera. + +--- + +## 🎯 Action & State Space Representation (will update if needed) + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +**Please refer to the subfolder README.md for more details.** + +--- + +## ⏱️ Data Synchronization Approach + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +We use a time-synchronized multi-modal data collection framework from an accepted ICRA 2026 paper: +``` +@inproceedings{zhou2026surgsync, + title={SurgSync: Time-Synchronized Multi-modal Data Collection Framework and Dataset for Surgical Robotics}, + author={Zhou, Haoying and ... and Kazanzides, Peter}, + booktitle={IEEE Intl. Conf. on Robotics and Automation (ICRA)}, + year={2026} +} +``` +We have two desktop: one for dVRK running and one for all the other pipelines. They are connected using ROS network and its NTP. + +We have two modes when data collection, and the performance is highly dependent on the hardware. + +**Online(-matching) Recorder**: (not uploaded yet) + +The design enforces strict time synchronization using multi-threading: only samples that fall within a user-defined time tolerance are admitted. This yields slightly uneven inter-sample intervals (irregular dt), +but it retains the natural continuity of smooth teleoperation segments and avoids label/feature drift. In our study, a time tolerance of 10 ms is selected to ensure both high time +alignment tightness and consecutive recorder output. + +As a result, the recorder time latency is 6.36 (+/- 4.72) ms with a median of 5.58ms. The recording frequency is 4.04 (+\- 1.69) Hz. We will assemble everything assume 10 FPS. + +**Offline(-matching) Recorder**: (already fully uploaded) + +Our offline-matching approach decouples recording from time alignments to maximize +the recording system efficiency. Therefore, this recorder produces synchronized datasets in two stages: (i) a lightweight +recorder logs camera streams to videos (pre-synced) and raw kinematic streams to binary files to disk with minimal processing; +(ii) an offline post-processing pipeline reconstructs a fixed-rate frame sequence and, for each frame, gathers the five +closest samples using nearest-timestamp lookup for subsequent interpolation. Compared to the online-matching recorder (which +pairs visual and kinematic data in real time), this two-stage design avoids tolerance-based dropping of data during capture +yielding a higher throughput and uniform intervals between synchronized data packets at the cost of requiring more storage +and substantial time for post-collection time-matching and interpolation. + +As a result, the recorder time latency is 1.35 (+/- 0.81) ms with a median of 1.33ms. And the recording frequency is a solid 10.00 FPS. + +**Note:** *Both recorders may have few outliners due to our suboptimal workstation hardware.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Dataset Lead** | `[Haoying Zhou (student lead), Peter Kazanzides (faculty PI)]` | +| **Institution** | `[Johns Hopkins University, Worcester Polytechnic Institute (Haoying Zhou)]` | +| **Contact Email** | `[hzhou6@wpi.edu, hzhou62@jh.edu, pkaz@jhu.edu]` | +| **Citation (BibTeX)** | 
@misc{[SurgSyncExt],
  author = {[Haoying Zhou, Peter Kazanzides, Chang Liu, Junlin Wu, Yimeng Wu, Hongjun Wu]},
  title = {[SurgSyncExt: Time-Synchronized Multi-Modal Dataset for Surgical Robotics]},
  year = {2025},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Surgical/jhu/lcsr/smarts/SurgSync-stitch-coldcut/P1/README.md b/Surgical/jhu/lcsr/smarts/SurgSync-stitch-coldcut/P1/README.md new file mode 100644 index 0000000000000000000000000000000000000000..7389f86a6837fad418bb54e5bb2435645bae7cc2 --- /dev/null +++ b/Surgical/jhu/lcsr/smarts/SurgSync-stitch-coldcut/P1/README.md @@ -0,0 +1,229 @@ +# SurgSync-stitch-coldcut P1 + +Canonical SMARTS leaf metadata README. + +- Canonical path: `Surgical/jhu/lcsr/smarts/SurgSync-stitch-coldcut/P1/` +- Legacy source mapping: `Surgical/jhu/lscr/smarts/offline_recorder_extracted/offline_data_part1`. +- This leaf is one canonical part of the broader JHU SMARTS dataset. + +The broader SMARTS dataset includes two sub-datasets: **offline_recorder** and **online_recorder**. + +--- + +## πŸ“‹ At a Glance + +*Provide a one-sentence summary of your dataset.* + +Teleoperated demonstrations of a da Vinci robot (dVRK-Si) performing multiple canonical tasks on ex-vivo tissue or table-top phantoms; ~50% data collected within the Intuitive abdominal dome model. All data collection uses a modern stereo chip-on-tip endoscope. Both endoscope camera calibration and robot hand-eye calibration will be provided. + +--- + +## File Structure + +For the dataset, it should + +```text +./offline_recorder or online_recorder +β”œβ”€β”€ calibration/ +β”‚ β”œβ”€β”€ case-*... +β”‚ β”‚ β”œβ”€β”€ camera calibration +β”‚ β”‚ β”‚ β”œβ”€β”€ left.yaml +β”‚ β”‚ β”‚ β”œβ”€β”€ right.yaml +β”‚ β”‚ β”‚ └── stereo_calib_params.json +β”‚ β”‚ └── hand_eye_calibration +β”‚ β”‚ β”‚ β”œβ”€β”€ PSM1/2-registration-dVRK.json +β”‚ β”‚ β”‚ └── PSM1/2-registration-open-cv.json +β”œβ”€β”€ data/ +β”‚ └── case-*... +β”œβ”€β”€ videos/ +β”‚ └── case-*... +β”œβ”€β”€ meta/ +β”‚ β”œβ”€β”€ episodes.jsonl +β”‚ β”œβ”€β”€ episodes_stats.jsonl +β”‚ β”œβ”€β”€ tasks.jsonl +β”‚ β”œβ”€β”€ info.json +β”‚ └── README.md +└── total_time.json +``` + +--- + +## πŸ“– Dataset Overview + +*Briefly describe the purpose and content of this dataset. What key skills or scenarios does it demonstrate?* + +This dataset comprises 2500+ teleoperated expert demonstrations of multiple canonical tasks on ex-vivo tissue or table-top phantoms using a surgical robotic system. The tasks include simple interrupted stitch, cold-cut dissection, peg transfer and tissue manipulation. It includes stereo endoscope calibration and robot hand–eye calibration (use dVRK camera registration format), along with high-fidelity, sharp stereo endoscopic video captured via a modern chip-on-tip endoscope. In addition to full task executions, the dataset provides consecutive subtask-level trajectories, enabling analysis of skill composition and procedural structure. Overall, it supports research in imitation learning and skill learning, surgical gesture/subtask recognition, and advanced perception (e.g., tool/tissue interaction understanding and visual tracking) in realistic surgical scenarios + +| | | +| :--- |:-------------------------------------------------------------------------------------------------------------------------------------------------| +| **Total Trajectories** | `1087 ex-vivo (offline) + 1061 ex-vivo (online) + 361 phantom (online)` | +| **Total Hours** | `2.83 ex-vivo (offline) + 3.31 ex-vivo (online) + 0.35 phantom (online)` | +| **Data Type** | `[ ] Clinical` `[X] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +**Note:** The user study experiments have been conducted under HIRB00000701 at Johns Hopkins University. + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* + +The primary skills or procedures demonstrated in this dataset include but not limited to: + +- simple interrupted stitching and its subtasks +- cold cut dissection and its subtasks +- peg transfer and its subtasks +- tissue manipulation and its subtasks +- ... + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Operator Count** | `[13]` | +| **Operator Skill Level** | `[4] Expert (e.g., Surgeon, Sonographer)`
`[5] Intermediate (e.g., Trained Researcher)`
`[4] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-08-01]` to `[2026-01-08]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [X] **No** + +**If yes, please briefly describe the recovery process:** + +**Note:** For your reference, this dataset includes recovery and failure demonstrations, but they are not labeled. + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [X] **Target Object** (e.g., different phantom models, suture types) +- [X] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [X] **Task Execution** (e.g., different techniques for the same task) +- [X] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +The camera (endoscope) can move with dVRK-Si ECM (endoscopy camera manipulator) as needed. The lighting conditions will be changed due to the camera movements. For the stitch, it has both ex-vivo tissue (chicken breast) and silicon phantom cases. The orientation of the wounds could vary from cases to cases. For cold-cut dissection, the target tissue can be different. The options include chicken drumsticks, thin-slide beef/pork, pork belly. The target suture needle could be in various locations. The background can be ex-vivo tissue or the surgical drapes. The detailed subtask approach can vary from cases to cases. For example, the knot-typing could be double throw or single throw. + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `dVRK-Si (the next generation da Vinci Reseach Kit)` + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- |:------------------------------------------------------------------------------------------------------------------------| +| **Primary Camera** | `Endoscopic Camera from Cornerstone Robotics Limited, 1920x1080 @ 60fps, recoreded in 10 FPS` | +| **Room/3rd Person Camera** | `Intel RealSense RGBD camera, only using RGB channel as a mono side-view camera, 1920x1080 @ 30fps, recorded in 10 FPS` | +| **Force/Torque Sensor** | `N/A` | +| **Medical Imager** | `N/A` | +| **Other** | `[Specify]` | + +**Note** The camera calibration files and hand-eye calibration matrices are just for the primary camera. + +--- + +## 🎯 Action & State Space Representation (will update if needed) + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +**Please refer to the subfolder README.md for more details.** + +--- + +## ⏱️ Data Synchronization Approach + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +We use a time-synchronized multi-modal data collection framework from an accepted ICRA 2026 paper: +``` +@inproceedings{zhou2026surgsync, + title={SurgSync: Time-Synchronized Multi-modal Data Collection Framework and Dataset for Surgical Robotics}, + author={Zhou, Haoying and ... and Kazanzides, Peter}, + booktitle={IEEE Intl. Conf. on Robotics and Automation (ICRA)}, + year={2026} +} +``` +We have two desktop: one for dVRK running and one for all the other pipelines. They are connected using ROS network and its NTP. + +We have two modes when data collection, and the performance is highly dependent on the hardware. + +**Online(-matching) Recorder**: (not uploaded yet) + +The design enforces strict time synchronization using multi-threading: only samples that fall within a user-defined time tolerance are admitted. This yields slightly uneven inter-sample intervals (irregular dt), +but it retains the natural continuity of smooth teleoperation segments and avoids label/feature drift. In our study, a time tolerance of 10 ms is selected to ensure both high time +alignment tightness and consecutive recorder output. + +As a result, the recorder time latency is 6.36 (+/- 4.72) ms with a median of 5.58ms. The recording frequency is 4.04 (+\- 1.69) Hz. We will assemble everything assume 10 FPS. + +**Offline(-matching) Recorder**: (already fully uploaded) + +Our offline-matching approach decouples recording from time alignments to maximize +the recording system efficiency. Therefore, this recorder produces synchronized datasets in two stages: (i) a lightweight +recorder logs camera streams to videos (pre-synced) and raw kinematic streams to binary files to disk with minimal processing; +(ii) an offline post-processing pipeline reconstructs a fixed-rate frame sequence and, for each frame, gathers the five +closest samples using nearest-timestamp lookup for subsequent interpolation. Compared to the online-matching recorder (which +pairs visual and kinematic data in real time), this two-stage design avoids tolerance-based dropping of data during capture +yielding a higher throughput and uniform intervals between synchronized data packets at the cost of requiring more storage +and substantial time for post-collection time-matching and interpolation. + +As a result, the recorder time latency is 1.35 (+/- 0.81) ms with a median of 1.33ms. And the recording frequency is a solid 10.00 FPS. + +**Note:** *Both recorders may have few outliners due to our suboptimal workstation hardware.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Dataset Lead** | `[Haoying Zhou (student lead), Peter Kazanzides (faculty PI)]` | +| **Institution** | `[Johns Hopkins University, Worcester Polytechnic Institute (Haoying Zhou)]` | +| **Contact Email** | `[hzhou6@wpi.edu, hzhou62@jh.edu, pkaz@jhu.edu]` | +| **Citation (BibTeX)** | 
@misc{[SurgSyncExt],
  author = {[Haoying Zhou, Peter Kazanzides, Chang Liu, Junlin Wu, Yimeng Wu, Hongjun Wu]},
  title = {[SurgSyncExt: Time-Synchronized Multi-Modal Dataset for Surgical Robotics]},
  year = {2025},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Surgical/jhu/lcsr/smarts/SurgSync-stitch-coldcut/P2/README.md b/Surgical/jhu/lcsr/smarts/SurgSync-stitch-coldcut/P2/README.md new file mode 100644 index 0000000000000000000000000000000000000000..9f492e0330ce25a043f0ef7f7336c4da57c815a0 --- /dev/null +++ b/Surgical/jhu/lcsr/smarts/SurgSync-stitch-coldcut/P2/README.md @@ -0,0 +1,229 @@ +# SurgSync-stitch-coldcut P2 + +Canonical SMARTS leaf metadata README. + +- Canonical path: `Surgical/jhu/lcsr/smarts/SurgSync-stitch-coldcut/P2/` +- Legacy source mapping: `Surgical/jhu/lscr/smarts/offline_recorder_extracted/offline_data_part2`. +- This leaf is one canonical part of the broader JHU SMARTS dataset. + +The broader SMARTS dataset includes two sub-datasets: **offline_recorder** and **online_recorder**. + +--- + +## πŸ“‹ At a Glance + +*Provide a one-sentence summary of your dataset.* + +Teleoperated demonstrations of a da Vinci robot (dVRK-Si) performing multiple canonical tasks on ex-vivo tissue or table-top phantoms; ~50% data collected within the Intuitive abdominal dome model. All data collection uses a modern stereo chip-on-tip endoscope. Both endoscope camera calibration and robot hand-eye calibration will be provided. + +--- + +## File Structure + +For the dataset, it should + +```text +./offline_recorder or online_recorder +β”œβ”€β”€ calibration/ +β”‚ β”œβ”€β”€ case-*... +β”‚ β”‚ β”œβ”€β”€ camera calibration +β”‚ β”‚ β”‚ β”œβ”€β”€ left.yaml +β”‚ β”‚ β”‚ β”œβ”€β”€ right.yaml +β”‚ β”‚ β”‚ └── stereo_calib_params.json +β”‚ β”‚ └── hand_eye_calibration +β”‚ β”‚ β”‚ β”œβ”€β”€ PSM1/2-registration-dVRK.json +β”‚ β”‚ β”‚ └── PSM1/2-registration-open-cv.json +β”œβ”€β”€ data/ +β”‚ └── case-*... +β”œβ”€β”€ videos/ +β”‚ └── case-*... +β”œβ”€β”€ meta/ +β”‚ β”œβ”€β”€ episodes.jsonl +β”‚ β”œβ”€β”€ episodes_stats.jsonl +β”‚ β”œβ”€β”€ tasks.jsonl +β”‚ β”œβ”€β”€ info.json +β”‚ └── README.md +└── total_time.json +``` + +--- + +## πŸ“– Dataset Overview + +*Briefly describe the purpose and content of this dataset. What key skills or scenarios does it demonstrate?* + +This dataset comprises 2500+ teleoperated expert demonstrations of multiple canonical tasks on ex-vivo tissue or table-top phantoms using a surgical robotic system. The tasks include simple interrupted stitch, cold-cut dissection, peg transfer and tissue manipulation. It includes stereo endoscope calibration and robot hand–eye calibration (use dVRK camera registration format), along with high-fidelity, sharp stereo endoscopic video captured via a modern chip-on-tip endoscope. In addition to full task executions, the dataset provides consecutive subtask-level trajectories, enabling analysis of skill composition and procedural structure. Overall, it supports research in imitation learning and skill learning, surgical gesture/subtask recognition, and advanced perception (e.g., tool/tissue interaction understanding and visual tracking) in realistic surgical scenarios + +| | | +| :--- |:-------------------------------------------------------------------------------------------------------------------------------------------------| +| **Total Trajectories** | `1087 ex-vivo (offline) + 1061 ex-vivo (online) + 361 phantom (online)` | +| **Total Hours** | `2.83 ex-vivo (offline) + 3.31 ex-vivo (online) + 0.35 phantom (online)` | +| **Data Type** | `[ ] Clinical` `[X] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +**Note:** The user study experiments have been conducted under HIRB00000701 at Johns Hopkins University. + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* + +The primary skills or procedures demonstrated in this dataset include but not limited to: + +- simple interrupted stitching and its subtasks +- cold cut dissection and its subtasks +- peg transfer and its subtasks +- tissue manipulation and its subtasks +- ... + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Operator Count** | `[13]` | +| **Operator Skill Level** | `[4] Expert (e.g., Surgeon, Sonographer)`
`[5] Intermediate (e.g., Trained Researcher)`
`[4] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-08-01]` to `[2026-01-08]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [X] **No** + +**If yes, please briefly describe the recovery process:** + +**Note:** For your reference, this dataset includes recovery and failure demonstrations, but they are not labeled. + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [X] **Target Object** (e.g., different phantom models, suture types) +- [X] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [X] **Task Execution** (e.g., different techniques for the same task) +- [X] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +The camera (endoscope) can move with dVRK-Si ECM (endoscopy camera manipulator) as needed. The lighting conditions will be changed due to the camera movements. For the stitch, it has both ex-vivo tissue (chicken breast) and silicon phantom cases. The orientation of the wounds could vary from cases to cases. For cold-cut dissection, the target tissue can be different. The options include chicken drumsticks, thin-slide beef/pork, pork belly. The target suture needle could be in various locations. The background can be ex-vivo tissue or the surgical drapes. The detailed subtask approach can vary from cases to cases. For example, the knot-typing could be double throw or single throw. + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `dVRK-Si (the next generation da Vinci Reseach Kit)` + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- |:------------------------------------------------------------------------------------------------------------------------| +| **Primary Camera** | `Endoscopic Camera from Cornerstone Robotics Limited, 1920x1080 @ 60fps, recoreded in 10 FPS` | +| **Room/3rd Person Camera** | `Intel RealSense RGBD camera, only using RGB channel as a mono side-view camera, 1920x1080 @ 30fps, recorded in 10 FPS` | +| **Force/Torque Sensor** | `N/A` | +| **Medical Imager** | `N/A` | +| **Other** | `[Specify]` | + +**Note** The camera calibration files and hand-eye calibration matrices are just for the primary camera. + +--- + +## 🎯 Action & State Space Representation (will update if needed) + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +**Please refer to the subfolder README.md for more details.** + +--- + +## ⏱️ Data Synchronization Approach + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +We use a time-synchronized multi-modal data collection framework from an accepted ICRA 2026 paper: +``` +@inproceedings{zhou2026surgsync, + title={SurgSync: Time-Synchronized Multi-modal Data Collection Framework and Dataset for Surgical Robotics}, + author={Zhou, Haoying and ... and Kazanzides, Peter}, + booktitle={IEEE Intl. Conf. on Robotics and Automation (ICRA)}, + year={2026} +} +``` +We have two desktop: one for dVRK running and one for all the other pipelines. They are connected using ROS network and its NTP. + +We have two modes when data collection, and the performance is highly dependent on the hardware. + +**Online(-matching) Recorder**: (not uploaded yet) + +The design enforces strict time synchronization using multi-threading: only samples that fall within a user-defined time tolerance are admitted. This yields slightly uneven inter-sample intervals (irregular dt), +but it retains the natural continuity of smooth teleoperation segments and avoids label/feature drift. In our study, a time tolerance of 10 ms is selected to ensure both high time +alignment tightness and consecutive recorder output. + +As a result, the recorder time latency is 6.36 (+/- 4.72) ms with a median of 5.58ms. The recording frequency is 4.04 (+\- 1.69) Hz. We will assemble everything assume 10 FPS. + +**Offline(-matching) Recorder**: (already fully uploaded) + +Our offline-matching approach decouples recording from time alignments to maximize +the recording system efficiency. Therefore, this recorder produces synchronized datasets in two stages: (i) a lightweight +recorder logs camera streams to videos (pre-synced) and raw kinematic streams to binary files to disk with minimal processing; +(ii) an offline post-processing pipeline reconstructs a fixed-rate frame sequence and, for each frame, gathers the five +closest samples using nearest-timestamp lookup for subsequent interpolation. Compared to the online-matching recorder (which +pairs visual and kinematic data in real time), this two-stage design avoids tolerance-based dropping of data during capture +yielding a higher throughput and uniform intervals between synchronized data packets at the cost of requiring more storage +and substantial time for post-collection time-matching and interpolation. + +As a result, the recorder time latency is 1.35 (+/- 0.81) ms with a median of 1.33ms. And the recording frequency is a solid 10.00 FPS. + +**Note:** *Both recorders may have few outliners due to our suboptimal workstation hardware.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Dataset Lead** | `[Haoying Zhou (student lead), Peter Kazanzides (faculty PI)]` | +| **Institution** | `[Johns Hopkins University, Worcester Polytechnic Institute (Haoying Zhou)]` | +| **Contact Email** | `[hzhou6@wpi.edu, hzhou62@jh.edu, pkaz@jhu.edu]` | +| **Citation (BibTeX)** | 
@misc{[SurgSyncExt],
  author = {[Haoying Zhou, Peter Kazanzides, Chang Liu, Junlin Wu, Yimeng Wu, Hongjun Wu]},
  title = {[SurgSyncExt: Time-Synchronized Multi-Modal Dataset for Surgical Robotics]},
  year = {2025},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Surgical/jhu/lcsr/smarts/SurgSync-stitch-coldcut/P3/README.md b/Surgical/jhu/lcsr/smarts/SurgSync-stitch-coldcut/P3/README.md new file mode 100644 index 0000000000000000000000000000000000000000..c3852c9fe40cf0fd851c6447a3a078bfbe0d9fc6 --- /dev/null +++ b/Surgical/jhu/lcsr/smarts/SurgSync-stitch-coldcut/P3/README.md @@ -0,0 +1,229 @@ +# SurgSync-stitch-coldcut P3 + +Canonical SMARTS leaf metadata README. + +- Canonical path: `Surgical/jhu/lcsr/smarts/SurgSync-stitch-coldcut/P3/` +- Legacy source mapping: `Surgical/jhu/lscr/smarts/offline_recorder_extracted/offline_data_part3`. +- This leaf is one canonical part of the broader JHU SMARTS dataset. + +The broader SMARTS dataset includes two sub-datasets: **offline_recorder** and **online_recorder**. + +--- + +## πŸ“‹ At a Glance + +*Provide a one-sentence summary of your dataset.* + +Teleoperated demonstrations of a da Vinci robot (dVRK-Si) performing multiple canonical tasks on ex-vivo tissue or table-top phantoms; ~50% data collected within the Intuitive abdominal dome model. All data collection uses a modern stereo chip-on-tip endoscope. Both endoscope camera calibration and robot hand-eye calibration will be provided. + +--- + +## File Structure + +For the dataset, it should + +```text +./offline_recorder or online_recorder +β”œβ”€β”€ calibration/ +β”‚ β”œβ”€β”€ case-*... +β”‚ β”‚ β”œβ”€β”€ camera calibration +β”‚ β”‚ β”‚ β”œβ”€β”€ left.yaml +β”‚ β”‚ β”‚ β”œβ”€β”€ right.yaml +β”‚ β”‚ β”‚ └── stereo_calib_params.json +β”‚ β”‚ └── hand_eye_calibration +β”‚ β”‚ β”‚ β”œβ”€β”€ PSM1/2-registration-dVRK.json +β”‚ β”‚ β”‚ └── PSM1/2-registration-open-cv.json +β”œβ”€β”€ data/ +β”‚ └── case-*... +β”œβ”€β”€ videos/ +β”‚ └── case-*... +β”œβ”€β”€ meta/ +β”‚ β”œβ”€β”€ episodes.jsonl +β”‚ β”œβ”€β”€ episodes_stats.jsonl +β”‚ β”œβ”€β”€ tasks.jsonl +β”‚ β”œβ”€β”€ info.json +β”‚ └── README.md +└── total_time.json +``` + +--- + +## πŸ“– Dataset Overview + +*Briefly describe the purpose and content of this dataset. What key skills or scenarios does it demonstrate?* + +This dataset comprises 2500+ teleoperated expert demonstrations of multiple canonical tasks on ex-vivo tissue or table-top phantoms using a surgical robotic system. The tasks include simple interrupted stitch, cold-cut dissection, peg transfer and tissue manipulation. It includes stereo endoscope calibration and robot hand–eye calibration (use dVRK camera registration format), along with high-fidelity, sharp stereo endoscopic video captured via a modern chip-on-tip endoscope. In addition to full task executions, the dataset provides consecutive subtask-level trajectories, enabling analysis of skill composition and procedural structure. Overall, it supports research in imitation learning and skill learning, surgical gesture/subtask recognition, and advanced perception (e.g., tool/tissue interaction understanding and visual tracking) in realistic surgical scenarios + +| | | +| :--- |:-------------------------------------------------------------------------------------------------------------------------------------------------| +| **Total Trajectories** | `1087 ex-vivo (offline) + 1061 ex-vivo (online) + 361 phantom (online)` | +| **Total Hours** | `2.83 ex-vivo (offline) + 3.31 ex-vivo (online) + 0.35 phantom (online)` | +| **Data Type** | `[ ] Clinical` `[X] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +**Note:** The user study experiments have been conducted under HIRB00000701 at Johns Hopkins University. + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* + +The primary skills or procedures demonstrated in this dataset include but not limited to: + +- simple interrupted stitching and its subtasks +- cold cut dissection and its subtasks +- peg transfer and its subtasks +- tissue manipulation and its subtasks +- ... + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Operator Count** | `[13]` | +| **Operator Skill Level** | `[4] Expert (e.g., Surgeon, Sonographer)`
`[5] Intermediate (e.g., Trained Researcher)`
`[4] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-08-01]` to `[2026-01-08]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [X] **No** + +**If yes, please briefly describe the recovery process:** + +**Note:** For your reference, this dataset includes recovery and failure demonstrations, but they are not labeled. + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [X] **Target Object** (e.g., different phantom models, suture types) +- [X] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [X] **Task Execution** (e.g., different techniques for the same task) +- [X] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +The camera (endoscope) can move with dVRK-Si ECM (endoscopy camera manipulator) as needed. The lighting conditions will be changed due to the camera movements. For the stitch, it has both ex-vivo tissue (chicken breast) and silicon phantom cases. The orientation of the wounds could vary from cases to cases. For cold-cut dissection, the target tissue can be different. The options include chicken drumsticks, thin-slide beef/pork, pork belly. The target suture needle could be in various locations. The background can be ex-vivo tissue or the surgical drapes. The detailed subtask approach can vary from cases to cases. For example, the knot-typing could be double throw or single throw. + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `dVRK-Si (the next generation da Vinci Reseach Kit)` + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- |:------------------------------------------------------------------------------------------------------------------------| +| **Primary Camera** | `Endoscopic Camera from Cornerstone Robotics Limited, 1920x1080 @ 60fps, recoreded in 10 FPS` | +| **Room/3rd Person Camera** | `Intel RealSense RGBD camera, only using RGB channel as a mono side-view camera, 1920x1080 @ 30fps, recorded in 10 FPS` | +| **Force/Torque Sensor** | `N/A` | +| **Medical Imager** | `N/A` | +| **Other** | `[Specify]` | + +**Note** The camera calibration files and hand-eye calibration matrices are just for the primary camera. + +--- + +## 🎯 Action & State Space Representation (will update if needed) + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +**Please refer to the subfolder README.md for more details.** + +--- + +## ⏱️ Data Synchronization Approach + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +We use a time-synchronized multi-modal data collection framework from an accepted ICRA 2026 paper: +``` +@inproceedings{zhou2026surgsync, + title={SurgSync: Time-Synchronized Multi-modal Data Collection Framework and Dataset for Surgical Robotics}, + author={Zhou, Haoying and ... and Kazanzides, Peter}, + booktitle={IEEE Intl. Conf. on Robotics and Automation (ICRA)}, + year={2026} +} +``` +We have two desktop: one for dVRK running and one for all the other pipelines. They are connected using ROS network and its NTP. + +We have two modes when data collection, and the performance is highly dependent on the hardware. + +**Online(-matching) Recorder**: (not uploaded yet) + +The design enforces strict time synchronization using multi-threading: only samples that fall within a user-defined time tolerance are admitted. This yields slightly uneven inter-sample intervals (irregular dt), +but it retains the natural continuity of smooth teleoperation segments and avoids label/feature drift. In our study, a time tolerance of 10 ms is selected to ensure both high time +alignment tightness and consecutive recorder output. + +As a result, the recorder time latency is 6.36 (+/- 4.72) ms with a median of 5.58ms. The recording frequency is 4.04 (+\- 1.69) Hz. We will assemble everything assume 10 FPS. + +**Offline(-matching) Recorder**: (already fully uploaded) + +Our offline-matching approach decouples recording from time alignments to maximize +the recording system efficiency. Therefore, this recorder produces synchronized datasets in two stages: (i) a lightweight +recorder logs camera streams to videos (pre-synced) and raw kinematic streams to binary files to disk with minimal processing; +(ii) an offline post-processing pipeline reconstructs a fixed-rate frame sequence and, for each frame, gathers the five +closest samples using nearest-timestamp lookup for subsequent interpolation. Compared to the online-matching recorder (which +pairs visual and kinematic data in real time), this two-stage design avoids tolerance-based dropping of data during capture +yielding a higher throughput and uniform intervals between synchronized data packets at the cost of requiring more storage +and substantial time for post-collection time-matching and interpolation. + +As a result, the recorder time latency is 1.35 (+/- 0.81) ms with a median of 1.33ms. And the recording frequency is a solid 10.00 FPS. + +**Note:** *Both recorders may have few outliners due to our suboptimal workstation hardware.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Dataset Lead** | `[Haoying Zhou (student lead), Peter Kazanzides (faculty PI)]` | +| **Institution** | `[Johns Hopkins University, Worcester Polytechnic Institute (Haoying Zhou)]` | +| **Contact Email** | `[hzhou6@wpi.edu, hzhou62@jh.edu, pkaz@jhu.edu]` | +| **Citation (BibTeX)** | 
@misc{[SurgSyncExt],
  author = {[Haoying Zhou, Peter Kazanzides, Chang Liu, Junlin Wu, Yimeng Wu, Hongjun Wu]},
  title = {[SurgSyncExt: Time-Synchronized Multi-Modal Dataset for Surgical Robotics]},
  year = {2025},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Surgical/obuda/frs_dome_1/README.md b/Surgical/obuda/frs_dome_1/README.md new file mode 100644 index 0000000000000000000000000000000000000000..458d8e34ac3ecf22eb54fdb97bc8e932f6e8a8ab --- /dev/null +++ b/Surgical/obuda/frs_dome_1/README.md @@ -0,0 +1,214 @@ + + +# README + +--- + +## πŸ“‹ At a Glance + +Teleoperated demonstrations of a da Vinci robot performing knot tying and suturing tasks on the FRS Dome phantom. + +--- + +## πŸ“– Dataset Overview + +This dataset contains trajectories of a laparoscopic surgeon using the dVRK to perform knot tying and suturing practice tasks on the FRS Dome (phantom info: https://www.surgicalexcellence.org/frs-dome). It includes successful trials, failures, and recovery attempts as well. + +| | | +| :--- | :--- | +| **Total Trajectories** | `102` | +| **Total Hours** | `01:18:23` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- Knot tying +- Suturing (stitching) + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `3` | +| **Operator Skill Level** | `[X] Expert (Laparoscopic surgeon)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[X] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2026-01-27]` to `[2026-01-28]` | + +### Recovery Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + + +- Knot tying: + - Re-grasping of the thread + - Thread stuck on the tool but recovered + +- Suturing: not distinguished + + +### Failure Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + + +- Knot tying: + - Thread stuck on tool + +- Suturing: not distinguished + + +--- + +## πŸ’‘ Diversity Dimensions + + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [ ] **Target Object** +- [ ] **Spatial Layout** +- [ ] **Robot Embodiment** +- [ ] **Task Execution** +- [X] **Background / Scene** +- [X] **Other** (Please specify: `Setup joints`, `Thread length`, `Technique`) + + + +Details: + +- Endoscope lighting was changed throughout the trials (all tasks) +- Natural and ceiling background light changed (all tasks) +- Camera positon was varied (all tasks) +- Set up joint configuration was varied (all tasks) +- Thread length was varied (suturing) +- Different stitching techniques (suturing) + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + + +- **Robot 1:** da Vinci Classic (with da Vinci Research Kit) + +### Sensors & Cameras + + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Stereo Endoscopic Camera, 720x576 @ 30fps` | +| **Wrist Camera** | `Endoscopic Camera (x2), 640x480 @ 30fps` | +| **Realsense Camera** | `Realsense RGB + D Camera, 1280x720 @ 30fps` | + +--- + +## 🎯 Action & State Space Representation + + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** + +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [x] **Other** (Please specify: `[Set Up Joints Configuration], [Camera Pose]`) + +**State Dimensions:** + +``` +observation.state: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### πŸ“‹ Additional Representations + +--- + +## ⏱️ Data Synchronization Approach + + +*Each modality (DeckLink cameras, USB cameras, and robotic kinematics) was recorded time-stamped on the same PC. A post-processing synchronization script segmented trials using explicit start/end markers and used the kinematic time series as the reference timeline. For each kinematic timestamp, the temporally nearest image frame from each camera stream was selected. The original stereoendoscope of the system is capable of only ~20FPS, thus the <50ms delay can not always be ensured.* + +--- + +## πŸ‘₯ Attribution & Contact + + +| | | +| :--- | :--- | +| **Dataset Lead** | `[KristΓ³f TakΓ‘cs, Eszter LukΓ‘cs, KristΓ³f MΓ³ga, TamΓ‘s Haidegger]` | +| **Institution** | `[Obuda University]` | +| **Contact Email** | `[krsitof.takacs@irob.uni-obuda.hu, eszter.lukacs@irob.uni-obuda.hu, kristof.moga@irob.uni-obuda.hu, haidegger@irob.uni-obuda.hu]` | +| **Citation (BibTeX)** | | diff --git a/Surgical/obuda/needlethreading_1/README.md b/Surgical/obuda/needlethreading_1/README.md new file mode 100644 index 0000000000000000000000000000000000000000..552377c2b98e760ec15864edcf3434b091b4c935 --- /dev/null +++ b/Surgical/obuda/needlethreading_1/README.md @@ -0,0 +1,201 @@ + + +# README + +--- + +## πŸ“‹ At a Glance + +Teleoperated demonstrations with a da Vinci robot performing the "needle threading" surgical practice task. + +--- + +## πŸ“– Dataset Overview + +This dataset contains trajectories of novices using the dVRK to perform the "needle threading" surgical practice task. It includes successful trials, failures, and recovery attempts as well. One episode is defined as the threading of a string through the loop of an eye bolt. + +| | | +| :--- | :--- | +| **Total Trajectories** | `196` | +| **Total Hours** | `00:57:16` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- String threading + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `3` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[X] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2026-01-23]` to `[2026-01-26]` | + +### Recovery Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + +- The string misses the ye of the bolt on the 1st attempt +- String is not grasped on 1st attempt during handover + + +### Failure Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the failures:** + +- Tool collision +- Pushing the board by hitting the bolts with a tool +- Repeatedly missing the eye of the bolt + +--- + +## πŸ’‘ Diversity Dimensions + + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [ ] **Target Object** (e.g., different phantom models, suture types) +- [X] **Spatial Layout** (Varying starting positions and board placement ) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [X] **Background / Scene** (different colors in the background) +- [X] **Other** (Please specify: `[Setup joints]`) + +*If you checked any of the above please briefly elaborate below.* + +- Endoscope lighting was changed throughout the trials +- Camera positon was varied +- Set up joint configuration was varied +- The board and the string position at start were varied + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +- **Robot 1:** da Vinci Classic (with da Vinci Research Kit) + +### Sensors & Cameras + + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Stereo Endoscopic Camera, 720x576 @ 30fps` | +| **Wrist Camera** | `Endoscopic Camera (x2), 640x480 @ 30fps` | +| **Realsense Camera** | `Realsense RGB + D Camera, 1280x720 @ 30fps` | + +--- + +## 🎯 Action & State Space Representation + + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** + +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [x] **Other** (Please specify: `[Set Up Joints Configuration], [Camera Pose]`) + +**State Dimensions:** + +``` +observation.state: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### πŸ“‹ Additional Representations + +--- + +## ⏱️ Data Synchronization Approach + + +*Each modality (DeckLink cameras, USB cameras, and robotic kinematics) was recorded time-stamped on the same PC. A post-processing synchronization script segmented trials using explicit start/end markers and used the kinematic time series as the reference timeline. For each kinematic timestamp, the temporally nearest image frame from each camera stream was selected. The original stereoendoscope of the system is capable of only ~20FPS, thus the <50ms delay can not always be ensured.* + +--- + +## πŸ‘₯ Attribution & Contact + + +| | | +| :--- | :--- | +| **Dataset Lead** | `[KristΓ³f TakΓ‘cs, Eszter LukΓ‘cs, TamΓ‘s Haidegger]` | +| **Institution** | `[Obuda University]` | +| **Contact Email** | `[krsitof.takacs@irob.uni-obuda.hu, eszter.lukacs@irob.uni-obuda.hu, haidegger@irob.uni-obuda.hu]` | +| **Citation (BibTeX)** | | diff --git a/Surgical/obuda/needlethreading_2/README.md b/Surgical/obuda/needlethreading_2/README.md new file mode 100644 index 0000000000000000000000000000000000000000..635031f761266aa1eea490d5899719fb1eacef8a --- /dev/null +++ b/Surgical/obuda/needlethreading_2/README.md @@ -0,0 +1,202 @@ + + +# README + +--- + +## πŸ“‹ At a Glance + +Teleoperated demonstrations with a da Vinci robot performing the "needle threading" surgical practice task. + +--- + +## πŸ“– Dataset Overview + +This dataset contains trajectories of novices using the dVRK to perform the "needle threading" surgical practice task. It includes successful trials, failures, and recovery attempts as well. One episode is defined as the threading of a string through the loop of an eye bolt. + +| | | +| :--- | :--- | +| **Total Trajectories** | `204` | +| **Total Hours** | `00:56:47` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- String threading + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `2` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[X] Intermediate (e.g., Trained Researcher)`
`[X] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2026-02-10]` to `[2026-02-11]` | + +### Recovery Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + +- The string misses the eye of the bolt on the 1st attempt +- String is not grasped on 1st attempt during handover + + +### Failure Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the failures:** + +- Tool collision +- Pushing the board by hitting the bolts with a tool +- Repeatedly missing the eye of the bolt + +--- + +## πŸ’‘ Diversity Dimensions + + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [ ] **Target Object** (e.g., different phantom models, suture types) +- [X] **Spatial Layout** (Varying starting positions and board placement ) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [X] **Background / Scene** (different colors in the background) +- [X] **Other** (Please specify: `[Setup joints]`) + +*If you checked any of the above please briefly elaborate below.* + +- Endoscope lighting was changed throughout the trials +- Camera positon was varied +- Set up joint configuration was varied +- The board and the string position at start were varied +- Background was varied + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +- **Robot 1:** da Vinci Classic (with da Vinci Research Kit) + +### Sensors & Cameras + + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Stereo Endoscopic Camera, 720x576 @ 30fps` | +| **Wrist Camera** | `Endoscopic Camera (x2), 640x480 @ 30fps` | +| **Realsense Camera** | `Realsense RGB + D Camera, 1280x720 @ 30fps` | + +--- + +## 🎯 Action & State Space Representation + + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** + +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [x] **Other** (Please specify: `[Set Up Joints Configuration], [Camera Pose]`) + +**State Dimensions:** + +``` +observation.state: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### πŸ“‹ Additional Representations + +--- + +## ⏱️ Data Synchronization Approach + + +*Each modality (DeckLink cameras, USB cameras, and robotic kinematics) was recorded time-stamped on the same PC. A post-processing synchronization script segmented trials using explicit start/end markers and used the kinematic time series as the reference timeline. For each kinematic timestamp, the temporally nearest image frame from each camera stream was selected. The original stereoendoscope of the system is capable of only ~20FPS, thus the <50ms delay can not always be ensured.* + +--- + +## πŸ‘₯ Attribution & Contact + + +| | | +| :--- | :--- | +| **Dataset Lead** | `[KristΓ³f TakΓ‘cs, Eszter LukΓ‘cs, TamΓ‘s Haidegger]` | +| **Institution** | `[Obuda University]` | +| **Contact Email** | `[krsitof.takacs@irob.uni-obuda.hu, eszter.lukacs@irob.uni-obuda.hu, haidegger@irob.uni-obuda.hu]` | +| **Citation (BibTeX)** | | diff --git a/Surgical/obuda/pegtransfer_1/README.md b/Surgical/obuda/pegtransfer_1/README.md new file mode 100644 index 0000000000000000000000000000000000000000..944c49fd5f4660cd6f1cf7b925fabcd821826d3a --- /dev/null +++ b/Surgical/obuda/pegtransfer_1/README.md @@ -0,0 +1,206 @@ + + +# README + +--- + +## πŸ“‹ At a Glance + +Teleoperated demonstrations of a da Vinci robot performing peg transfer on a 3D printed model with silicone pegs. + +--- + +## πŸ“– Dataset Overview + +This dataset contains trajectories of novices using the dVRK to perform peg transfer. It includes successful trials, failures, and recovery attempts as well. One episode demonstrates the transfer of 1 peg. + +| | | +| :--- | :--- | +| **Total Trajectories** | `216` | +| **Total Hours** | `01:14:54` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- Peg transfer + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `3` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[X] Intermediate (e.g., Trained Researcher)`
`[X] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2026-01-22]` to `[2026-01-23]` | + +### Recovery Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + + +- The peg is not grasped accurately, the operator has to re-grasp the peg. +- Pegs are placed incorrectly onto the rods, the operator has to reposition. + + +### Failure Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + + +- The peg falls out of the grasp. +- The tools hit the board and push it extensively. + + +--- + +## πŸ’‘ Diversity Dimensions + + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [ ] **Target Object** +- [ ] **Spatial Layout** +- [ ] **Robot Embodiment** +- [ ] **Task Execution** +- [X] **Background / Scene** +- [X] **Other** (Please specify: `Setup joints`) + + + +Details: + +- Endoscope lighting was changed throughout the trials +- Natural and ceiling background light changed +- Camera positon was varied +- Set up joint configuration was varied + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + + +- **Robot 1:** da Vinci Classic (with da Vinci Research Kit) + +### Sensors & Cameras + + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Stereo Endoscopic Camera, 720x576 @ 30fps` | +| **Wrist Camera** | `Endoscopic Camera (x2), 640x480 @ 30fps` | +| **Realsense Camera** | `Realsense RGB + D Camera, 1280x720 @ 30fps` | + +--- + +## 🎯 Action & State Space Representation + + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** + +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [x] **Other** (Please specify: `[Set Up Joints Configuration], [Camera Pose]`) + +**State Dimensions:** + +``` +observation.state: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### πŸ“‹ Additional Representations + +--- + +## ⏱️ Data Synchronization Approach + + +*Each modality (DeckLink cameras, USB cameras, and robotic kinematics) was recorded time-stamped on the same PC. A post-processing synchronization script segmented trials using explicit start/end markers and used the kinematic time series as the reference timeline. For each kinematic timestamp, the temporally nearest image frame from each camera stream was selected. The original stereoendoscope of the system is capable of only ~20FPS, thus the <50ms delay can not always be ensured.* + +--- + +## πŸ‘₯ Attribution & Contact + + +| | | +| :--- | :--- | +| **Dataset Lead** | `[KristΓ³f TakΓ‘cs, Eszter LukΓ‘cs, LΓ³rΓ‘nt Domokos, TamΓ‘s Haidegger]` | +| **Institution** | `[Obuda University]` | +| **Contact Email** | `[krsitof.takacs@irob.uni-obuda.hu, eszter.lukacs@irob.uni-obuda.hu, lorant.domokos@irob.uni-obuda.hu, haidegger@irob.uni-obuda.hu]` | +| **Citation (BibTeX)** | | diff --git a/Surgical/obuda/pegtransfer_2/README.md b/Surgical/obuda/pegtransfer_2/README.md new file mode 100644 index 0000000000000000000000000000000000000000..962da4b54d3bd2ebad6a018f38bf1f7d10698bec --- /dev/null +++ b/Surgical/obuda/pegtransfer_2/README.md @@ -0,0 +1,206 @@ + + +# README + +--- + +## πŸ“‹ At a Glance + +Teleoperated demonstrations of a da Vinci robot performing peg transfer on a 3D printed model with silicone pegs. + +--- + +## πŸ“– Dataset Overview + +This dataset contains trajectories of novices using the dVRK to perform peg transfer. It includes successful trials, failures, and recovery attempts as well. One episode demonstrates the transfer of 1 peg. + +| | | +| :--- | :--- | +| **Total Trajectories** | `184` | +| **Total Hours** | `00:43:25` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- Peg transfer + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `2` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[X] Intermediate (e.g., Trained Researcher)`
`[X] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2026-01-22]` to `[2026-01-23]` | + +### Recovery Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + + +- The peg is not grasped accurately, the operator has to re-grasp the peg. +- Pegs are placed incorrectly onto the rods, the operator has to reposition. + + +### Failure Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + + +- The peg falls out of the grasp. +- The tools hit the board and push it extensively. + + +--- + +## πŸ’‘ Diversity Dimensions + + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [ ] **Target Object** +- [ ] **Spatial Layout** +- [ ] **Robot Embodiment** +- [ ] **Task Execution** +- [X] **Background / Scene** +- [X] **Other** (Please specify: `Setup joints`) + + + +Details: + +- Endoscope lighting was changed throughout the trials +- Natural and ceiling background light changed +- Camera positon was varied +- Set up joint configuration was varied + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + + +- **Robot 1:** da Vinci Classic (with da Vinci Research Kit) + +### Sensors & Cameras + + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Stereo Endoscopic Camera, 720x576 @ 30fps` | +| **Wrist Camera** | `Endoscopic Camera (x2), 640x480 @ 30fps` | +| **Realsense Camera** | `Realsense RGB + D Camera, 1280x720 @ 30fps` | + +--- + +## 🎯 Action & State Space Representation + + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** + +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [x] **Other** (Please specify: `[Set Up Joints Configuration], [Camera Pose]`) + +**State Dimensions:** + +``` +observation.state: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### πŸ“‹ Additional Representations + +--- + +## ⏱️ Data Synchronization Approach + + +*Each modality (DeckLink cameras, USB cameras, and robotic kinematics) was recorded time-stamped on the same PC. A post-processing synchronization script segmented trials using explicit start/end markers and used the kinematic time series as the reference timeline. For each kinematic timestamp, the temporally nearest image frame from each camera stream was selected. The original stereoendoscope of the system is capable of only ~20FPS, thus the <50ms delay can not always be ensured.* + +--- + +## πŸ‘₯ Attribution & Contact + + +| | | +| :--- | :--- | +| **Dataset Lead** | `[KristΓ³f TakΓ‘cs, Eszter LukΓ‘cs, LΓ³rΓ‘nt Domokos, TamΓ‘s Haidegger]` | +| **Institution** | `[Obuda University]` | +| **Contact Email** | `[krsitof.takacs@irob.uni-obuda.hu, eszter.lukacs@irob.uni-obuda.hu, lorant.domokos@irob.uni-obuda.hu, haidegger@irob.uni-obuda.hu]` | +| **Citation (BibTeX)** | | diff --git a/Surgical/obuda/pork_1/README.md b/Surgical/obuda/pork_1/README.md new file mode 100644 index 0000000000000000000000000000000000000000..87ac8c1663102f1f936d43cced17f5bca2fd4024 --- /dev/null +++ b/Surgical/obuda/pork_1/README.md @@ -0,0 +1,204 @@ + + +# README + +--- + +## πŸ“‹ At a Glance + +Teleoperated demonstrations of cutting small tissue samples from a fresh pork shoulder using the da Vinci robot. + +--- + +## πŸ“– Dataset Overview + +This dataset contains trajectories of novice operators using the dVRK to excise small tissue samples from fresh porcine shoulder. It includes successful trials, failures, and recovery attempts. One episode consists of tissue grasping, complete excision, and lifting of the resected sample. + +| | | +| :--- | :--- | +| **Total Trajectories** | `318` | +| **Total Hours** | `01:31:56` | +| **Data Type** | `[ ] Clinical` `[X] Ex-Vivo` `[ ] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- Tissue grasping +- Tissue excision + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `3` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[X] Intermediate (e.g., Trained Researcher)`
`[X] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2026-02-11]` to `[2026-02-13]` | + +### Recovery Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + + +- The tissue has to be re-grasped during the episode + + +### Failure Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + + +- The tissue tears at an unintended location away from the intended cutting line. + +--- + +## πŸ’‘ Diversity Dimensions + + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [ ] **Target Object** +- [ ] **Spatial Layout** +- [ ] **Robot Embodiment** +- [ ] **Task Execution** +- [ ] **Background / Scene** +- [X] **Other** (Please specify: `Setup joints`) + + + +Details: + +- Endoscope lighting was changed throughout the trials +- Natural and ceiling background light changed +- Camera positon was varied +- Set up joint configuration was varied + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + + +- **Robot 1:** da Vinci Classic (with da Vinci Research Kit) + +### Sensors & Cameras + + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Stereo Endoscopic Camera, 720x576 @ 30fps` | +| **Wrist Camera** | `Endoscopic Camera (x2), 640x480 @ 30fps` | +| **Realsense Camera** | `Realsense RGB + D Camera, 1280x720 @ 30fps` | + +--- + +## 🎯 Action & State Space Representation + + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** + +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [x] **Other** (Please specify: `[Set Up Joints Configuration], [Camera Pose]`) + +**State Dimensions:** + +``` +observation.state: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### πŸ“‹ Additional Representations + +--- + +## ⏱️ Data Synchronization Approach + + +*Each modality (DeckLink cameras, USB cameras, and robotic kinematics) was recorded time-stamped on the same PC. A post-processing synchronization script segmented trials using explicit start/end markers and used the kinematic time series as the reference timeline. For each kinematic timestamp, the temporally nearest image frame from each camera stream was selected. The original stereoendoscope of the system is capable of only ~20FPS, thus the <50ms delay can not always be ensured.* + +--- + +## πŸ‘₯ Attribution & Contact + + +| | | +| :--- | :--- | +| **Dataset Lead** | `[KristΓ³f TakΓ‘cs, Eszter LukΓ‘cs, TamΓ‘s Haidegger]` | +| **Institution** | `[Obuda University]` | +| **Contact Email** | `[krsitof.takacs@irob.uni-obuda.hu, eszter.lukacs@irob.uni-obuda.hu, haidegger@irob.uni-obuda.hu]` | +| **Citation (BibTeX)** | | diff --git a/Surgical/obuda/rollercoaster_1/README.md b/Surgical/obuda/rollercoaster_1/README.md new file mode 100644 index 0000000000000000000000000000000000000000..fb4a276b9a39a9379a05eb15931b5aa7ffd52e67 --- /dev/null +++ b/Surgical/obuda/rollercoaster_1/README.md @@ -0,0 +1,199 @@ + + +# README + +--- + +## πŸ“‹ At a Glance + +Teleoperated demonstrations with a da Vinci robot performing the "rollercoaster" (or "hot wire") surgical practice task. + +--- + +## πŸ“– Dataset Overview + +This dataset contains trajectories of novices using the dVRK to perform the "rollercoaster" surgical practice task. It includes successful trials, failures, and recovery attempts as well. One episode lasts from grasping the ring until releasing it at the other end of the wire. + +| | | +| :--- | :--- | +| **Total Trajectories** | `95` | +| **Total Hours** | `01:12:22` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- Spatial navigation + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `3` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[X] Intermediate (e.g., Trained Researcher)`
`[X] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2026-01-28]` to `[2026-01-29]` | + +### Recovery Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + +- The ring needs to be re-grasped or re-aligned + + +### Failure Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the failures:** + +- Tool collision +- Extensively moving the board + + +--- + +## πŸ’‘ Diversity Dimensions + + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [ ] **Target Object** (e.g., different phantom models, suture types) +- [X] **Spatial Layout** (Varying starting positions and board placement ) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [X] **Background / Scene** (different colors in the background) +- [X] **Other** (Please specify: `[Setup joints]`) + +*If you checked any of the above please briefly elaborate below.* + +- Endoscope lighting was changed throughout the trials +- Camera positon was varied +- Set up joint configuration was varied +- The board's position and orientation were varied + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +- **Robot 1:** da Vinci Classic (with da Vinci Research Kit) + +### Sensors & Cameras + + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Stereo Endoscopic Camera, 720x576 @ 30fps` | +| **Wrist Camera** | `Endoscopic Camera (x2), 640x480 @ 30fps` | +| **Realsense Camera** | `Realsense RGB + D Camera, 1280x720 @ 30fps` | + +--- + +## 🎯 Action & State Space Representation + + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** + +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [x] **Other** (Please specify: `[Set Up Joints Configuration], [Camera Pose]`) + +**State Dimensions:** + +``` +observation.state: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### πŸ“‹ Additional Representations + +--- + +## ⏱️ Data Synchronization Approach + + +*Each modality (DeckLink cameras, USB cameras, and robotic kinematics) was recorded time-stamped on the same PC. A post-processing synchronization script segmented trials using explicit start/end markers and used the kinematic time series as the reference timeline. For each kinematic timestamp, the temporally nearest image frame from each camera stream was selected. The original stereoendoscope of the system is capable of only ~20FPS, thus the <50ms delay can not always be ensured.* + +--- + +## πŸ‘₯ Attribution & Contact + + +| | | +| :--- | :--- | +| **Dataset Lead** | `[KristΓ³f TakΓ‘cs, Eszter LukΓ‘cs, LΓ³rΓ‘nt Domokos, TamΓ‘s Haidegger]` | +| **Institution** | `[Obuda University]` | +| **Contact Email** | `[krsitof.takacs@irob.uni-obuda.hu, eszter.lukacs@irob.uni-obuda.hu, lorant.domokos@irob.uni-obuda.hu, haidegger@irob.uni-obuda.hu]` | +| **Citation (BibTeX)** | | diff --git a/Surgical/obuda/seaspike_1/README.md b/Surgical/obuda/seaspike_1/README.md new file mode 100644 index 0000000000000000000000000000000000000000..0a47dff4d7ac0deb04522e36acdf531a616d3e24 --- /dev/null +++ b/Surgical/obuda/seaspike_1/README.md @@ -0,0 +1,200 @@ + + +# README + +--- + +## πŸ“‹ At a Glance + +Teleoperated demonstrations with a da Vinci robot performing the "seaspike" surgical practice task. + +--- + +## πŸ“– Dataset Overview + +This dataset contains trajectories of novices using the dVRK to perform the "seaspike" surgical practice task. It includes successful trials, failures, and recovery attempts as well. One episode lasts from picking up a ring until placing it on a spike. + +| | | +| :--- | :--- | +| **Total Trajectories** | `207` | +| **Total Hours** | `00:49:36` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- Ring transfer + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `2` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[X] Intermediate (e.g., Trained Researcher)`
`[X] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2026-01-28]` to `[2026-01-29]` | + +### Recovery Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + +- The ring has to be re-grasped at pickup +- The ring misses the peak of the spike for the first try, but is not dropped down + + +### Failure Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the failures:** + +- The ring is misplaced, dropped down + +--- + +## πŸ’‘ Diversity Dimensions + + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [ ] **Target Object** (e.g., different phantom models, suture types) +- [X] **Spatial Layout** (Varying starting positions and board placement ) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [X] **Background / Scene** (different colors in the background) +- [X] **Other** (Please specify: `[Setup joints]`) + +*If you checked any of the above please briefly elaborate below.* + +- Endoscope lighting was changed throughout the trials +- Camera positon was varied +- Set up joint configuration was varied +- Background changes +- The board's orientation (i.e., the position of the spikes) was varied + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +- **Robot 1:** da Vinci Classic (with da Vinci Research Kit) + +### Sensors & Cameras + + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Stereo Endoscopic Camera, 720x576 @ 30fps` | +| **Wrist Camera** | `Endoscopic Camera (x2), 640x480 @ 30fps` | +| **Realsense Camera** | `Realsense RGB + D Camera, 1280x720 @ 30fps` | + +--- + +## 🎯 Action & State Space Representation + + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** + +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [x] **Other** (Please specify: `[Set Up Joints Configuration], [Camera Pose]`) + +**State Dimensions:** + +``` +observation.state: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### πŸ“‹ Additional Representations + +--- + +## ⏱️ Data Synchronization Approach + + +*Each modality (DeckLink cameras, USB cameras, and robotic kinematics) was recorded time-stamped on the same PC. A post-processing synchronization script segmented trials using explicit start/end markers and used the kinematic time series as the reference timeline. For each kinematic timestamp, the temporally nearest image frame from each camera stream was selected. The original stereoendoscope of the system is capable of only ~20FPS, thus the <50ms delay can not always be ensured.* + +--- + +## πŸ‘₯ Attribution & Contact + + +| | | +| :--- | :--- | +| **Dataset Lead** | `[KristΓ³f TakΓ‘cs, Eszter LukΓ‘cs, LΓ³rΓ‘nt Domokos, TamΓ‘s Haidegger]` | +| **Institution** | `[Obuda University]` | +| **Contact Email** | `[krsitof.takacs@irob.uni-obuda.hu, eszter.lukacs@irob.uni-obuda.hu, lorant.domokos@irob.uni-obuda.hu, haidegger@irob.uni-obuda.hu]` | +| **Citation (BibTeX)** | | diff --git a/Surgical/obuda/seaspike_2/README.md b/Surgical/obuda/seaspike_2/README.md new file mode 100644 index 0000000000000000000000000000000000000000..7df2012a2b41a2dbf0434789297bddc7982851f2 --- /dev/null +++ b/Surgical/obuda/seaspike_2/README.md @@ -0,0 +1,200 @@ + + +# README + +--- + +## πŸ“‹ At a Glance + +Teleoperated demonstrations with a da Vinci robot performing the "seaspike" surgical practice task. + +--- + +## πŸ“– Dataset Overview + +This dataset contains trajectories of novices using the dVRK to perform the "seaspike" surgical practice task. It includes successful trials, failures, and recovery attempts as well. One episode lasts from picking up a ring until placing it on a spike. + +| | | +| :--- | :--- | +| **Total Trajectories** | `153` | +| **Total Hours** | `00:37:35` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- Ring transfer + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `2` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[X] Intermediate (e.g., Trained Researcher)`
`[X] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2026-02-04]` to `[2026-02-06]` | + +### Recovery Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + +- The ring has to be re-grasped at pickup +- The ring misses the peak of the spike for the first try, but is not dropped down + + +### Failure Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the failures:** + +- The ring is misplaced, dropped down + +--- + +## πŸ’‘ Diversity Dimensions + + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [ ] **Target Object** (e.g., different phantom models, suture types) +- [X] **Spatial Layout** (Varying starting positions and board placement ) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [X] **Background / Scene** (different colors in the background) +- [X] **Other** (Please specify: `[Setup joints]`) + +*If you checked any of the above please briefly elaborate below.* + +- Endoscope lighting was changed throughout the trials +- Camera positon was varied +- Set up joint configuration was varied +- Background changes +- The board's orientation (i.e., the position of the spikes) was varied + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +- **Robot 1:** da Vinci Classic (with da Vinci Research Kit) + +### Sensors & Cameras + + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Stereo Endoscopic Camera, 720x576 @ 30fps` | +| **Wrist Camera** | `Endoscopic Camera (x2), 640x480 @ 30fps` | +| **Realsense Camera** | `Realsense RGB + D Camera, 1280x720 @ 30fps` | + +--- + +## 🎯 Action & State Space Representation + + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** + +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [x] **Other** (Please specify: `[Set Up Joints Configuration], [Camera Pose]`) + +**State Dimensions:** + +``` +observation.state: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### πŸ“‹ Additional Representations + +--- + +## ⏱️ Data Synchronization Approach + + +*Each modality (DeckLink cameras, USB cameras, and robotic kinematics) was recorded time-stamped on the same PC. A post-processing synchronization script segmented trials using explicit start/end markers and used the kinematic time series as the reference timeline. For each kinematic timestamp, the temporally nearest image frame from each camera stream was selected. The original stereoendoscope of the system is capable of only ~20FPS, thus the <50ms delay can not always be ensured.* + +--- + +## πŸ‘₯ Attribution & Contact + + +| | | +| :--- | :--- | +| **Dataset Lead** | `[KristΓ³f TakΓ‘cs, Eszter LukΓ‘cs, LΓ³rΓ‘nt Domokos, TamΓ‘s Haidegger]` | +| **Institution** | `[Obuda University]` | +| **Contact Email** | `[krsitof.takacs@irob.uni-obuda.hu, eszter.lukacs@irob.uni-obuda.hu, lorant.domokos@irob.uni-obuda.hu, haidegger@irob.uni-obuda.hu]` | +| **Citation (BibTeX)** | | diff --git a/Surgical/obuda/seaspike_3/README.md b/Surgical/obuda/seaspike_3/README.md new file mode 100644 index 0000000000000000000000000000000000000000..c5b06a521f879436303945cc42be6ab04ad1acc0 --- /dev/null +++ b/Surgical/obuda/seaspike_3/README.md @@ -0,0 +1,200 @@ + + +# README + +--- + +## πŸ“‹ At a Glance + +Teleoperated demonstrations with a da Vinci robot performing the "seaspike" surgical practice task. + +--- + +## πŸ“– Dataset Overview + +This dataset contains trajectories of novices using the dVRK to perform the "seaspike" surgical practice task. It includes successful trials, failures, and recovery attempts as well. One episode lasts from picking up a ring until placing it on a spike. + +| | | +| :--- | :--- | +| **Total Trajectories** | `219` | +| **Total Hours** | `00:57:12` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- Ring transfer + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `2` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[X] Intermediate (e.g., Trained Researcher)`
`[X] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2026-02-06]` to `[2026-02-09]` | + +### Recovery Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + +- The ring has to be re-grasped at pickup +- The ring misses the peak of the spike for the first try, but is not dropped down + + +### Failure Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the failures:** + +- The ring is misplaced, dropped down + +--- + +## πŸ’‘ Diversity Dimensions + + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [ ] **Target Object** (e.g., different phantom models, suture types) +- [X] **Spatial Layout** (Varying starting positions and board placement ) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [X] **Background / Scene** (different colors in the background) +- [X] **Other** (Please specify: `[Setup joints]`) + +*If you checked any of the above please briefly elaborate below.* + +- Endoscope lighting was changed throughout the trials +- Camera positon was varied +- Set up joint configuration was varied +- Background changes +- The board's orientation (i.e., the position of the spikes) was varied + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +- **Robot 1:** da Vinci Classic (with da Vinci Research Kit) + +### Sensors & Cameras + + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Stereo Endoscopic Camera, 720x576 @ 30fps` | +| **Wrist Camera** | `Endoscopic Camera (x2), 640x480 @ 30fps` | +| **Realsense Camera** | `Realsense RGB + D Camera, 1280x720 @ 30fps` | + +--- + +## 🎯 Action & State Space Representation + + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** + +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [x] **Other** (Please specify: `[Set Up Joints Configuration], [Camera Pose]`) + +**State Dimensions:** + +``` +observation.state: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### πŸ“‹ Additional Representations + +--- + +## ⏱️ Data Synchronization Approach + + +*Each modality (DeckLink cameras, USB cameras, and robotic kinematics) was recorded time-stamped on the same PC. A post-processing synchronization script segmented trials using explicit start/end markers and used the kinematic time series as the reference timeline. For each kinematic timestamp, the temporally nearest image frame from each camera stream was selected. The original stereoendoscope of the system is capable of only ~20FPS, thus the <50ms delay can not always be ensured.* + +--- + +## πŸ‘₯ Attribution & Contact + + +| | | +| :--- | :--- | +| **Dataset Lead** | `[KristΓ³f TakΓ‘cs, Eszter LukΓ‘cs, LΓ³rΓ‘nt Domokos, TamΓ‘s Haidegger]` | +| **Institution** | `[Obuda University]` | +| **Contact Email** | `[krsitof.takacs@irob.uni-obuda.hu, eszter.lukacs@irob.uni-obuda.hu, lorant.domokos@irob.uni-obuda.hu, haidegger@irob.uni-obuda.hu]` | +| **Citation (BibTeX)** | | diff --git a/Surgical/obuda/skinphantom_1/README.md b/Surgical/obuda/skinphantom_1/README.md new file mode 100644 index 0000000000000000000000000000000000000000..8eae77090fcb5becad6bb25d4dd142962fd87496 --- /dev/null +++ b/Surgical/obuda/skinphantom_1/README.md @@ -0,0 +1,217 @@ + + +# README + +--- + +## πŸ“‹ At a Glance + +Teleoperated demonstrations of a da Vinci robot performing interrupted suturing on a skin phantom, including needle driving and knot tying. + +--- + +## πŸ“– Dataset Overview + +This dataset contains teleoperated trajectories of an expert operator (da Vinci certified surgeon) performing wound closure on a skin phantom using the da Vinci Robot. The 4 subtasks capture the execution of a single interrupted stitch, including needle insertion (subtask 0), complete suture pull-through (subtask 1), and knot tying consisting of one surgeon’s knot (subtask 2) followed by two securing throws (subtask 3). The dataset includes successful trials, failure cases, and recovery attempts too. + +| | | +| :--- | :--- | +| **Total Trajectories** | `106` | +| **Total Hours** | `00:23:19` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[X] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- Needle driving +- Tissue approximation +- Knot tying + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + + +- [X] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `1` | +| **Operator Skill Level** | `[X] Expert (da Vinci certified surgeon)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | `[2026-02-11]` | + +### Recovery Demonstrations + + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** + +- Needle Insertion: + - Loss of needle grip (needle dropped) + - Needle fails to pass through the tissue on the first attempt + +- Suture Pulling: + - Misgrasping the suture + + - Single and Double Knot: + - Misgrasping the suture + - Suture became caught on the tool, but was successfully freed + - Dropping the suture + +### Failure Demonstrations + +- [X] **Yes** +- [ ] **No** + +**If yes, please briefly describe the failure process:** + +- Needle Insertion: not distinguished + +- Suture Pulling: not distinguished + +- Single and Double Knot: + - Knot was tied directly onto the suture + + +--- + +## πŸ’‘ Diversity Dimensions + + +- [X] **Camera Position / Angle** +- [X] **Lighting Conditions** +- [ ] **Target Object** +- [ ] **Spatial Layout** +- [ ] **Robot Embodiment** +- [ ] **Task Execution** +- [ ] **Background / Scene** +- [X] **Other** (Please specify: `Thread length`) + + + +Details: + +- Endoscope lighting was changed throughout the trials +- Natural and ceiling background light changed +- Camera positon was varied +- Thread length varied + + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + + +- **Robot 1:** da Vinci Classic (with da Vinci Research Kit) + +### Sensors & Cameras + + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Stereo Endoscopic Camera, 720x576 @ 30fps` | +| **Wrist Camera** | `Endoscopic Camera (x2), 640x480 @ 30fps` | +| **Realsense Camera** | `Realsense RGB + D Camera, 1280x720 @ 30fps` | + +--- + +## 🎯 Action & State Space Representation + + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** + +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [x] **Other** (Please specify: `[Set Up Joints Configuration], [Camera Pose]`) + +**State Dimensions:** + +``` +observation.state: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### πŸ“‹ Additional Representations + +--- + +## ⏱️ Data Synchronization Approach + + +*Each modality (DeckLink cameras, USB cameras, and robotic kinematics) was recorded time-stamped on the same PC. A post-processing synchronization script segmented trials using explicit start/end markers and used the kinematic time series as the reference timeline. For each kinematic timestamp, the temporally nearest image frame from each camera stream was selected. The original stereoendoscope of the system is capable of only ~20FPS, thus the <50ms delay can not always be ensured.* + +--- + +## πŸ‘₯ Attribution & Contact + + +| | | +| :--- | :--- | +| **Dataset Lead** | `[KristΓ³f TakΓ‘cs, Eszter LukΓ‘cs, LΓ‘szlΓ³ Piros, TamΓ‘s Haidegger]` | +| **Institution** | `[Obuda University]` | +| **Contact Email** | `[krsitof.takacs@irob.uni-obuda.hu, eszter.lukacs@irob.uni-obuda.hu, haidegger@irob.uni-obuda.hu]` | +| **Citation (BibTeX)** | | diff --git a/Surgical/semaphor/open_h_semaphor/README.md b/Surgical/semaphor/open_h_semaphor/README.md new file mode 100644 index 0000000000000000000000000000000000000000..ac0b040cf649e755e7bc35cd9df4a7c5a470c758 --- /dev/null +++ b/Surgical/semaphor/open_h_semaphor/README.md @@ -0,0 +1,175 @@ + + +# [Dataset Name] - README + +--- + +## πŸ“‹ At a Glance + +*Our dataset is dataset with tracked neural surgery tools doing suturing on ex-vivo pork belly* + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 500 trajectories of expert surgeons using neural surgery tool to perform surgical suturing tasks. It includes successful trials to provide a robust dataset for training imitation learning policies* + +| | | +| :--- | :--- | +| **Total Trajectories** | `534` | +| **Total Hours** | `0.5 hours` | +| **Data Type** | `[ ] Clinical` `[ X ] Ex-Vivo` `[ ] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [X] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +- Needle-passing +- Needle-Grasping +- Needle-handover +- Suture-tying + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [X] **Other** (Please specify: `Direct Human Operation`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[X] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-01-10]` to `[2025-01-28]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [X] **No** + +**If yes, please briefly describe the recovery process:** + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [ ] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [ ] **Target Object** (e.g., different phantom models, suture types) +- [ ] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*Manual Neural Surgery Tool* + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `[Zed X camera, 1920x1080 @ 30fps]` | +| **Side Camera (left)** | `[Zed X camera, 1920x1080 @ 30fps]` | +| **Side Camera (right)** | `[Zed X camera, 1920x1080 @ 30fps]` | + +--- + +## 🎯 Action & State Space Representation + +### Action Space Representation + +**Primary Action Representation:** +- [X] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [ ] **Quaternions** (x, y, z, w) +- [X] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [X] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*Action is represented as the pose of (x_0,y_0,z_0,roll_0,pitch_0,yall_0,valid_0,x_1,y_1,z_1,roll_1,pitch_1,yall_1,valid_1) with subscripts as tool id. valid means whether this pose for the tool is valid in this frame* + + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [X] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** + +*State is represented as the pose of (x_0,y_0,z_0,roll_0,pitch_0,yall_0,valid_0,x_1,y_1,z_1,roll_1,pitch_1,yall_1,valid_1) with subscripts as tool id. valid means whether this pose for the tool is valid in this frame.* + + +--- + +## ⏱️ Data Synchronization Approach + +*We have the timestamp from all cameras, we set the main camera as anchor, for each frame in the main camera, each side cameras choose the frame with closest timestamp as the symchromnized frame.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Hao Ding]` | +| **Institution** | `[Semaphor Surgical]` | +| **Contact Email** | `[hao@semaphorsurgical.com]` | +| **Citation (BibTeX)** | 
@misc{[SemaphorSuture],
  author = {[Hao Ding, Chenhao Yu, Chenhao Yu, Zoe SoulΓ©, Jose Porras, Axel Krieger, Mathias Unberath]},
  title = {[Semaphor Suturing Dataset]},
  year = {2025},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Surgical/stanford/collaborative_haptics_and_robotics_in_medicine_lab/real_robot_dvrk/needle_transfer/README.md b/Surgical/stanford/collaborative_haptics_and_robotics_in_medicine_lab/real_robot_dvrk/needle_transfer/README.md new file mode 100644 index 0000000000000000000000000000000000000000..cd7ab0856cc4e1ac7dba8bd33fc404b20e994ec6 --- /dev/null +++ b/Surgical/stanford/collaborative_haptics_and_robotics_in_medicine_lab/real_robot_dvrk/needle_transfer/README.md @@ -0,0 +1,232 @@ + + +# Needle Transfer - README + +--- + +## πŸ“‹ At a Glance + +*Teleoperated demonstrations of the da Vinci Si robot performing needle transfer with a suturing needle.* + + + +--- + +## πŸ“– Dataset Description + + +The dataset comprises 700 dVRK trajectories of needle transfer performed on a table top phantom, including successful trials, failures, and recovery attempts. It provides synchronized cartesian, joint and video data for training and evaluating robot learning policies. NOTE: The initial few episodes are a little zoomed out but the positioning of the camera is much closer after 100 episodes. + + +| | | +| :--- | :--- | +| **Total Trajectories** | `700` | +| **Total Hours** | `2.9` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[x] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [x] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* +- Needle Pickup +- Needle Passing +- Needle Collection + + +--- + +## πŸ”¬ Collection Procedure + +### Collection Method + +*How was the data collected?* + +- [x] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `1` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[x] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-10-01]` to `[2025-01-15]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [x] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** +The dataset includes 25 recovery demonstrations and 75 failure demonstrations. In the failure cases, the robotic arm fails to achieve a grasp or drops while passing. In the recovery cases, the arm grasps the object with an incorrect orientation for passing, after which the operator re-orients the grasp before completing the pass. + + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [x] **Camera Position / Angle** +- [x] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* +The camera configuration was adjusted every 50–100 demonstrations by varying the setup height by Β±2 cm. In addition, the needle type and phantom base were changed periodically. Lighting conditions were varied between 60% and 100%. Each demonstration also features a slightly different needle pickup location. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `dVRK (da Vinci Research Kit)` + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Endoscopic Camera, 1920x1080 @ 30fps with both left and right video feed` | +| **Joint/Position Encoders** | | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to camera frame (ECM)) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [ ] **Quaternions** (x, y, z, w) +- [x] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [x] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* +``` +action: [PSM{i}_jaw, PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z, PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw] +- PSM{i} represents the arm (Could be PSM1 or PSM2) +- PSM{i}_jaw: Jaw angle in radians +- PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z: Absolute position in camera frame (ECM frame) +- PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw: Absolute Orientation as Euler Angles +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +**Example:** +``` +observation.state: [PSM{i}_joint_1, PSM{i}_joint_2, PSM{i}_joint_3, PSM{i}_joint_4, PSM{i}_joint_5, PSM{i}_joint_6, PSM{i}_jaw, PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z, PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw] +- PSM{i} represents the arm (Could be PSM1 or PSM2) +- PSM{i}_joint_1 to PSM{i}_joint_6: Absolute joint positions for the 7-DOF arm (radians) +- PSM{i}_jaw: Jaw angle of the gripper (radians) +- PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z: End-effector absolute position in camera (ECM) frame (meters) +- PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw: End-effector absolute orientation as Euler angles (radians) + +``` + +--- + +## ⏱️ Data Synchronization Approach + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +We use the ApproximateTimeSynchronizer [[Link]](https://wiki.ros.org/message_filters/ApproximateTime) from the ROS message_filters package to synchronize all data streams. The queue_size parameter controls the number of incoming messages buffered for each topic, while the slop parameter specifies the maximum allowable time difference between messages for them to be considered synchronized. This approach aligns messages based on their timestamps within a defined tolerance rather than requiring exact matches. + +Data is recorded at 30 Hz, with the camera feed acting as the bottleneck. During data collection, we monitor the inter-frame time difference and ensure it remains close to 33 ms, resulting in approximately 450 frames per 15-second episode. In rare cases, message delays lead to significantly fewer frames (fewer than 435); such episodes are discarded and re-recorded. + + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Creators** | `Alaa Eldin Abdelaal`, `Chetan Reddy Narayanaswamy`, `Jiaqi Shao`, `Howard Ji`, `Allison Okamura` | +| **Institution** | `Stanford University` | +| **Contact Email** | `abdelaal@stanford.edu OR alaaaldinmagdy@gmail.com`, `chetanrn@stanford.edu`, `jiaqis7@stanford.edu`, `howardji@stanford.edu`, `aokamura@stanford.edu` | +| **Point of Contact** | `abdelaal@stanford.edu OR alaaaldinmagdy@gmail.com` | diff --git a/Surgical/stanford/collaborative_haptics_and_robotics_in_medicine_lab/real_robot_dvrk/peg_transfer/README.md b/Surgical/stanford/collaborative_haptics_and_robotics_in_medicine_lab/real_robot_dvrk/peg_transfer/README.md new file mode 100644 index 0000000000000000000000000000000000000000..abddf9e920ab8b14e8b15fe048e0b5f862f3e84f --- /dev/null +++ b/Surgical/stanford/collaborative_haptics_and_robotics_in_medicine_lab/real_robot_dvrk/peg_transfer/README.md @@ -0,0 +1,232 @@ + + +# Peg Transfer - README + +--- + +## πŸ“‹ At a Glance + +*Teleoperated demonstrations of the da Vinci Si robot performing peg transfer on a transfer board* + + + +--- + +## πŸ“– Dataset Description + + +The dataset comprises around 600 dVRK trajectories of the peg transfer task including successful trials, failures, and recovery attempts. It provides synchronized cartesian, joint and video data for training and evaluating robot learning policies. + + +| | | +| :--- | :--- | +| **Total Trajectories** | `598` | +| **Total Hours** | `2.5` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[x] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [x] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* +- Peg Pickup +- Peg Passing +- Peg Placing + + +--- + +## πŸ”¬ Collection Procedure + +### Collection Method + +*How was the data collected?* + +- [x] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `2` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[x] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-10-01]` to `[2025-01-15]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [x] **Yes** +- [ ] **No** + +**If yes, please briefly describe the recovery process:** +The dataset includes 50 recovery demonstrations and 50 failure demonstrations. In the failure cases, the robotic arm fails to achieve a grasp or drops while passing. In the recovery cases, the arm fails to pick the peg correctly but picks it up when attempted again. Another type of recovery occurs when the peg is being placed, fails and its orientation is adjusted to keep it upright. + + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [x] **Camera Position / Angle** +- [x] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* +The camera configuration was adjusted every 50–100 demonstrations by varying the setup height by Β±2 cm. In addition, the needle type and phantom base were changed periodically. Lighting conditions were varied between 60% and 100%. Each demonstration also features a slightly different needle pickup location. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `dVRK (da Vinci Research Kit)` + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Endoscopic Camera, 1920x1080 @ 30fps with both left and right video feed` | +| **Joint/Position Encoders** | | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to camera frame (ECM)) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [ ] **Quaternions** (x, y, z, w) +- [x] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [x] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* +``` +action: [PSM{i}_jaw, PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z, PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw] +- PSM{i} represents the arm (Could be PSM1 or PSM2) +- PSM{i}_jaw: Jaw angle in radians +- PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z: Absolute position in camera frame (ECM frame) +- PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw: Absolute Orientation as Euler Angles +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +**Example:** +``` +observation.state: [PSM{i}_joint_1, PSM{i}_joint_2, PSM{i}_joint_3, PSM{i}_joint_4, PSM{i}_joint_5, PSM{i}_joint_6, PSM{i}_jaw, PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z, PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw] +- PSM{i} represents the arm (Could be PSM1 or PSM2) +- PSM{i}_joint_1 to PSM{i}_joint_6: Absolute joint positions for the 7-DOF arm (radians) +- PSM{i}_jaw: Jaw angle of the gripper (radians) +- PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z: End-effector absolute position in camera (ECM) frame (meters) +- PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw: End-effector absolute orientation as Euler angles (radians) + +``` + +--- + +## ⏱️ Data Synchronization Approach + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +We use the ApproximateTimeSynchronizer [[Link]](https://wiki.ros.org/message_filters/ApproximateTime) from the ROS message_filters package to synchronize all data streams. The queue_size parameter controls the number of incoming messages buffered for each topic, while the slop parameter specifies the maximum allowable time difference between messages for them to be considered synchronized. This approach aligns messages based on their timestamps within a defined tolerance rather than requiring exact matches. + +Data is recorded at 30 Hz, with the camera feed acting as the bottleneck. During data collection, we monitor the inter-frame time difference and ensure it remains close to 33 ms, resulting in approximately 450 frames per 15-second episode. In rare cases, message delays lead to significantly fewer frames (fewer than 435); such episodes are discarded and re-recorded. + + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Creators** | `Alaa Eldin Abdelaal`, `Chetan Reddy Narayanaswamy`, `Jiaqi Shao`, `Howard Ji`, `Allison Okamura` | +| **Institution** | `Stanford University` | +| **Contact Email** | `abdelaal@stanford.edu OR alaaaldinmagdy@gmail.com`, `chetanrn@stanford.edu`, `jiaqis7@stanford.edu`, `howardji@stanford.edu`, `aokamura@stanford.edu` | +| **Point of Contact** | `abdelaal@stanford.edu OR alaaaldinmagdy@gmail.com` | diff --git a/Surgical/stanford/collaborative_haptics_and_robotics_in_medicine_lab/real_robot_dvrk/tissue_retraction/README.md b/Surgical/stanford/collaborative_haptics_and_robotics_in_medicine_lab/real_robot_dvrk/tissue_retraction/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b4b5dc2d7868798c8eea3ecb9c7a547b7ca1594f --- /dev/null +++ b/Surgical/stanford/collaborative_haptics_and_robotics_in_medicine_lab/real_robot_dvrk/tissue_retraction/README.md @@ -0,0 +1,230 @@ + + +# Tissue Retraction - README + +--- + +## πŸ“‹ At a Glance + +*Teleoperated demonstrations of the da Vinci Si robot performing tissue retraction of 2-3 layers (of silicone phantom)* + + + +--- + +## πŸ“– Dataset Description + + +The dataset comprises 700 dVRK trajectories of tissue retraction performed on a table top phantom, including successful trials, failures, and recovery attempts. It provides synchronized cartesian, joint and video data for training and evaluating robot learning policies. + + +| | | +| :--- | :--- | +| **Total Trajectories** | `698` | +| **Total Hours** | `2.9` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[x] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [x] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* +- Choosing Right Point to Start Retracting +- Second Arm Assisting +- Retracting Multiple Layers + + +--- + +## πŸ”¬ Collection Procedure + +### Collection Method + +*How was the data collected?* + +- [x] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `1` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[x] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-10-01]` to `[2025-01-15]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [x] **Camera Position / Angle** +- [x] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* +The camera configuration was adjusted every 50–100 demonstrations by varying the setup height by Β±2 cm. In addition, the needle type and phantom base were changed periodically. Lighting conditions were varied between 60% and 100%. Some demonstrations had 2 layers and some had 3 layers. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `dVRK (da Vinci Research Kit)` + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Endoscopic Camera, 1920x1080 @ 30fps with both left and right video feed` | +| **Joint/Position Encoders** | | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to camera frame (ECM)) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [ ] **Quaternions** (x, y, z, w) +- [x] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [x] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* +``` +action: [PSM{i}_jaw, PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z, PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw] +- PSM{i} represents the arm (Could be PSM1 or PSM2) +- PSM{i}_jaw: Jaw angle in radians +- PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z: Absolute position in camera frame (ECM frame) +- PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw: Absolute Orientation as Euler Angles +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +**Example:** +``` +observation.state: [PSM{i}_joint_1, PSM{i}_joint_2, PSM{i}_joint_3, PSM{i}_joint_4, PSM{i}_joint_5, PSM{i}_joint_6, PSM{i}_jaw, PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z, PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw] +- PSM{i} represents the arm (Could be PSM1 or PSM2) +- PSM{i}_joint_1 to PSM{i}_joint_6: Absolute joint positions for the 7-DOF arm (radians) +- PSM{i}_jaw: Jaw angle of the gripper (radians) +- PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z: End-effector absolute position in camera (ECM) frame (meters) +- PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw: End-effector absolute orientation as Euler angles (radians) + +``` + +--- + +## ⏱️ Data Synchronization Approach + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +We use the ApproximateTimeSynchronizer [[Link]](https://wiki.ros.org/message_filters/ApproximateTime) from the ROS message_filters package to synchronize all data streams. The queue_size parameter controls the number of incoming messages buffered for each topic, while the slop parameter specifies the maximum allowable time difference between messages for them to be considered synchronized. This approach aligns messages based on their timestamps within a defined tolerance rather than requiring exact matches. + +Data is recorded at 30 Hz, with the camera feed acting as the bottleneck. During data collection, we monitor the inter-frame time difference and ensure it remains close to 33 ms, resulting in approximately 450 frames per 15-second episode. In rare cases, message delays lead to significantly fewer frames (fewer than 435); such episodes are discarded and re-recorded. + + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Creators** | `Alaa Eldin Abdelaal`, `Chetan Reddy Narayanaswamy`, `Jiaqi Shao`, `Howard Ji`, `Allison Okamura` | +| **Institution** | `Stanford University` | +| **Contact Email** | `abdelaal@stanford.edu OR alaaaldinmagdy@gmail.com`, `chetanrn@stanford.edu`, `jiaqis7@stanford.edu`, `howardji@stanford.edu`, `aokamura@stanford.edu` | +| **Point of Contact** | `abdelaal@stanford.edu OR alaaaldinmagdy@gmail.com` | + diff --git a/Surgical/stanford/collaborative_haptics_and_robotics_in_medicine_lab/simulation/block_transfer_sim_lerobot_1_28/README.md b/Surgical/stanford/collaborative_haptics_and_robotics_in_medicine_lab/simulation/block_transfer_sim_lerobot_1_28/README.md new file mode 100644 index 0000000000000000000000000000000000000000..6c30e8e3e08d1a6964236e5d4805b7a76251d0c6 --- /dev/null +++ b/Surgical/stanford/collaborative_haptics_and_robotics_in_medicine_lab/simulation/block_transfer_sim_lerobot_1_28/README.md @@ -0,0 +1,228 @@ + + +# Block Transfer (Simulation) - README + +--- + +## πŸ“‹ At a Glance + +*Teleoperated demonstrations of the da Vinci robot in orbitsurgical library in Isaac Sim performing block transfer* + + + +--- + +## πŸ“– Dataset Description + + +The dataset comprises around 500 dVRK trajectories of the block transfer task including successful trials and some failure cases. It provides synchronized cartesian, joint and video data for training and evaluating robot learning policies. The asset used is from orbitsurgical library. + + +| | | +| :--- | :--- | +| **Total Trajectories** | `500` | +| **Total Hours** | `2.5` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[ ] Table-Top Phantom` `[X] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [x] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* +- Needle Pickup +- Needle Passing + + +--- + +## πŸ”¬ Collection Procedure + +### Collection Method + +*How was the data collected?* + +- [x] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `1` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[x] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-10-01]` to `[2026-01-15]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [X] **No** + +**If yes, please briefly describe the recovery process:** + + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [ ] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [ ] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [x] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* +Each demonstration has a slightly different needle pickup location and pass location, also with the different pick up and pass style. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `dVRK (da Vinci Research Kit)` + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Camera in Isaac Sim, 640x480 @ 20fps with both left and right video feed` | +| **Joint/Position Encoders** | | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to camera frame (ECM)) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [ ] **Quaternions** (x, y, z, w) +- [x] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* +``` +action: [PSM{i}_jaw, PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z, PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw] +- PSM{i} represents the arm (Could be PSM1 or PSM2) +- PSM{i}_jaw: Jaw angle in radians +- PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z: Absolute position in camera frame (ECM frame) +- PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw: Absolute Orientation as Euler Angles +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +``` +observation.state: [PSM{i}_joint_1, PSM{i}_joint_2, PSM{i}_joint_3, PSM{i}_joint_4, PSM{i}_joint_5, PSM{i}_joint_6, PSM{i}_jaw, PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z, PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw] +- PSM{i} represents the arm (Could be PSM1 or PSM2) +- PSM{i}_joint_1 to PSM{i}_joint_6: Absolute joint positions for the 7-DOF arm (radians) +- PSM{i}_jaw: Jaw angle of the gripper (radians) +- PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z: End-effector absolute position in robot frame (meters) +- PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw: End-effector absolute orientation as Euler angles (radians) + +``` + +--- + +## ⏱️ Data Synchronization Approach + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +Our capture loop captures robot state, timestamp, and both camera frames in one tuple per frame and enqueues it. The ring buffer guarantees FIFO delivery so that logger thread dequeues the same tuple, writes the paired images, and records the synchronized data in exactly one row. +Data is recorded around 20 Hz, with the camera feed acting as the bottleneck. resulting in approximately 460 frames per 25-second episode. + + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Creators** | `Alaa Eldin Abdelaal`, `Chetan Reddy Narayanaswamy`, `Jiaqi Shao`, `Howard Ji`, `Allison Okamura` | +| **Institution** | `Stanford University` | +| **Contact Email** | `abdelaal@stanford.edu OR alaaaldinmagdy@gmail.com`, `chetanrn@stanford.edu`, `jiaqis7@stanford.edu`, `howardji@stanford.edu`, `aokamura@stanford.edu` | +| **Point of Contact** | `abdelaal@stanford.edu OR alaaaldinmagdy@gmail.com` | diff --git a/Surgical/stanford/collaborative_haptics_and_robotics_in_medicine_lab/simulation/needle_transfer_sim_lerobot_1_28/README.md b/Surgical/stanford/collaborative_haptics_and_robotics_in_medicine_lab/simulation/needle_transfer_sim_lerobot_1_28/README.md new file mode 100644 index 0000000000000000000000000000000000000000..778480ee48778256160d5ec0770ffc0291440992 --- /dev/null +++ b/Surgical/stanford/collaborative_haptics_and_robotics_in_medicine_lab/simulation/needle_transfer_sim_lerobot_1_28/README.md @@ -0,0 +1,228 @@ + + +# Needle Transfer (Simulation)- README + +--- + +## πŸ“‹ At a Glance + +*Teleoperated demonstrations of the da Vinci robot in orbitsurgical library in Isaac Sim performing needle transfer* + + + +--- + +## πŸ“– Dataset Description + + +The dataset comprises around 500 dVRK trajectories of the needle transfer task including successful trials and some failure cases. It provides synchronized cartesian, joint and video data for training and evaluating robot learning policies. The asset used is from orbitsurgical library. + + +| | | +| :--- | :--- | +| **Total Trajectories** | `500` | +| **Total Hours** | `2.5` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[ ] Table-Top Phantom` `[X] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [x] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*List the primary skills or procedures demonstrated in this dataset.* +- Needle Pickup +- Needle Passing + + +--- + +## πŸ”¬ Collection Procedure + +### Collection Method + +*How was the data collected?* + +- [x] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `1` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[x] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-10-01]` to `[2026-01-15]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [X] **No** + +**If yes, please briefly describe the recovery process:** + + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [ ] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [ ] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [x] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* +Each demonstration has a slightly different needle pickup location and pass location, also with the different pick up and pass style. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `dVRK (da Vinci Research Kit)` + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `Camera in Isaac Sim, 640x480 @ 20fps with both left and right video feed` | +| **Joint/Position Encoders** | | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to camera frame (ECM)) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [ ] **Quaternions** (x, y, z, w) +- [x] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* +``` +action: [PSM{i}_jaw, PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z, PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw] +- PSM{i} represents the arm (Could be PSM1 or PSM2) +- PSM{i}_jaw: Jaw angle in radians +- PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z: Absolute position in camera frame (ECM frame) +- PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw: Absolute Orientation as Euler Angles +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [x] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +``` +observation.state: [PSM{i}_joint_1, PSM{i}_joint_2, PSM{i}_joint_3, PSM{i}_joint_4, PSM{i}_joint_5, PSM{i}_joint_6, PSM{i}_jaw, PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z, PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw] +- PSM{i} represents the arm (Could be PSM1 or PSM2) +- PSM{i}_joint_1 to PSM{i}_joint_6: Absolute joint positions for the 7-DOF arm (radians) +- PSM{i}_jaw: Jaw angle of the gripper (radians) +- PSM{i}_ee_x, PSM{i}_ee_y, PSM{i}_ee_z: End-effector absolute position in robot frame (meters) +- PSM{i}_ee_roll, PSM{i}_ee_pitch, PSM{i}_ee_yaw: End-effector absolute orientation as Euler angles (radians) + +``` + +--- + +## ⏱️ Data Synchronization Approach + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +Our capture loop captures robot state, timestamp, and both camera frames in one tuple per frame and enqueues it. The ring buffer guarantees FIFO delivery so that logger thread dequeues the same tuple, writes the paired images, and records the synchronized data in exactly one row. +Data is recorded around 20 Hz, with the camera feed acting as the bottleneck. resulting in approximately 460 frames per 25-second episode. + + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Creators** | `Alaa Eldin Abdelaal`, `Chetan Reddy Narayanaswamy`, `Jiaqi Shao`, `Howard Ji`, `Allison Okamura` | +| **Institution** | `Stanford University` | +| **Contact Email** | `abdelaal@stanford.edu OR alaaaldinmagdy@gmail.com`, `chetanrn@stanford.edu`, `jiaqis7@stanford.edu`, `howardji@stanford.edu`, `aokamura@stanford.edu` | +| **Point of Contact** | `abdelaal@stanford.edu OR alaaaldinmagdy@gmail.com` | diff --git a/Surgical/turin/mitic_lerobot_ex_vivo/README.md b/Surgical/turin/mitic_lerobot_ex_vivo/README.md new file mode 100644 index 0000000000000000000000000000000000000000..d78facfdb1d3b51093ef708e64ed2c238bc75d97 --- /dev/null +++ b/Surgical/turin/mitic_lerobot_ex_vivo/README.md @@ -0,0 +1,160 @@ +# DVRK Suturing Subtasks Dataset - README + +--- + +## At a Glance + +Teleoperated demonstrations of a dVRK robot performing suturing subtasksk, like tissue lifting, needle insertion, needle extraction, knot tying on ex vivo porcine colon and stomach. + +--- + +## Dataset Overview + +| | | +| :--- | :--- | +| **Total Trajectories** | 800 | +| **Total Hours** | [TO be filled] | +| **Data Type** | Ex-Vivo | +| **License** | CC BY 4.0 | +| **Version** | 1.0 | + +--- + +## Tasks & Domain + +### Domain + +- [x] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- Bimanual manipulation +- Pick and place +- Needle handling +- Knot tying +- Soft and ex-vivo tissue manipulation + +--- + +## Data Collection Details + +### Collection Method + +- [x] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | 5 operators | +| **Operator Skill Level** | Expert (Surgeons) and Novice (ML researchers with minimal surgical experience) | +| **Collection Period** | [To be filled] | + +### Recovery Demonstrations + +- [x] **Yes** +- [] **No** + +For each task, recovery demonstrations and errors are recorded. + +--- + +## Diversity Dimensions + +- [x] **Target Object** (different wounds for suturing and different anatomical specimens) +- [] **Spatial Layout** +- [x] **Camera Position / Angle** (different position and orientations) +- [x] **Lighting Conditions** +- [ ] **Robot Embodiment** +- [x] **Task Execution** (different operators performed the tasks differently) +- [x] **Background / Scene** (different background conditions) + +We have recorded tasks both on porcine colon and stomach. Camera position and lighting condition were changed after every 20 episodes circa. We recruited 3 expert surgeons and 2 ML researchers that performed the tasks according to their expertise and own techniques. The background conditions change between colon and stomach episodes. + +--- + +## Equipment & Setup + +### Robotic Platform(s) + +- **Robot 1:** dVRK (da Vinci Research Kit) + +### Sensors & Cameras + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | Stereo Endoscope (left), 1920x1080 @ 30 fps | +| **Secondary Camera** | Stereo Endoscope (right), 1920x1080 @ 30 fps | + +--- + +## Action & State Space Representation + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) + +**Action Dimensions:** +``` +action: [psm1_x, psm1_y, psm1_z, psm1_qx, psm1_qy, psm1_qz, psm1_qw, + psm2_x, psm2_y, psm2_z, psm2_qx, psm2_qy, psm2_qz, psm2_qw] +- psm1_x, psm1_y, psm1_z: PSM1 absolute cartesian positions +- psm1_qx, psm1_qy, psm1_qz, psm1_qw: PSM1 absolute cartesian orientations +- psm2_x, psm2_y, psm2_z: PSM2 absolute cartesian positions +- psm2_qx, psm2_qy, psm2_qz, psm2_qw: PSM2 absolute cartesian orientations +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) + +**State Dimensions:** +``` +observation.state: [psm1_j1, psm1_j2, psm1_j3, psm1_j4, psm1_j5, psm1_j6, + psm2_j1, psm2_j2, psm2_j3, psm2_j4, psm2_j5, psm2_j6] +- psm1_j1-j6: PSM1 6-DOF joint positions (radians) +- psm2_j1-j6: PSM2 6-DOF joint positions (radians) +``` + +--- + +## Data Synchronization Approach + +All data are acquired using rosbag and all ROS messages are stamped with their header.stamp fields. +Data is synchronized offline post collection. The right camera serves as the reference timestamp source. For each frame acquired, only the (time) closest data is saved from robot kinematics, with a tolerance of 0.1 secs. + +**Synchronization Method:** +- Reference stream: `/decklink/right/image_raw/compressed` (right stereo camera) +- Tolerance: 0.1 seconds +- Framework: ROS (Robot Operating System) + +Messages from the following topics are synchronized: +- `/decklink/left/image_raw/compressed` - Left camera images +- `/decklink/right/image_raw/compressed` - Right camera images +- `/PSM1/measured_js` - PSM1 joint states +- `/PSM1/measured_cp` - PSM1 absolute cartesian position +- `/PSM2/measured_js` - PSM2 joint states +- `/PSM2/measured_cp` - PSM2 absolute cartesian position + +--- + +## Attribution & Contact + +| | | +| :--- | :--- | +| **Dataset Lead** | Matteo Pescio, Francesco Marzola, Luigi Muratore, Federica Barontini, Giovanni Distefano, Federico Lavagno, Giulio Dagnino, Alberto Arezzo | +| **Institution** | MITIC Lab - UniversitΓ  degli Studi di Torino | +| **Contact Email** | matteo.pescio@unito.it, francesco.marzola@unito.it, luigi.muratore@studenti.polito.it, federica.barontini@unito.it, giovanni.distefano@unito.it, federico.lavagno@unito.it, giulio.dagnino@unito.it, alberto.arezzo@unito.it | +| **Citation (BibTeX)** | 
@misc{dvrk_suturing_subtasks_2025,
  author = {Pescio, Matteo and Marzola, Francesco and Muratore, Luigi and Barontini, Federica and Distefano, Giovanni and Lavagno, Federico and Dagnino, Giulio and Arezzo, Alberto},
  title = {DVRK Suturing Subtasks Dataset},
  year = {2025},
  publisher = {Open-H-Embodiment},
}
| + +--- diff --git a/Surgical/turin/mitic_lerobot_plastic_pad/README.md b/Surgical/turin/mitic_lerobot_plastic_pad/README.md new file mode 100644 index 0000000000000000000000000000000000000000..4eb267bb3361bfbe8b8649991ff8232fb1cd4adb --- /dev/null +++ b/Surgical/turin/mitic_lerobot_plastic_pad/README.md @@ -0,0 +1,160 @@ +# DVRK Suturing Subtasks Dataset - README + +--- + +## At a Glance + +Teleoperated demonstrations of a dVRK robot performing suturing subtasksk, like tissue lifting, needle insertion, needle extraction, knot tying on plastic phantom. + +--- + +## Dataset Overview + +| | | +| :--- | :--- | +| **Total Trajectories** | 550 | +| **Total Hours** | [TO be filled] | +| **Data Type** | Ex-Vivo | +| **License** | CC BY 4.0 | +| **Version** | 1.0 | + +--- + +## Tasks & Domain + +### Domain + +- [x] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- Bimanual manipulation +- Pick and place +- Needle handling +- Knot tying +- Soft tissue manipulation + +--- + +## Data Collection Details + +### Collection Method + +- [x] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | 5 operators | +| **Operator Skill Level** | Expert (Surgeons) and Novice (ML researchers with minimal surgical experience) | +| **Collection Period** | [To be filled] | + +### Recovery Demonstrations + +- [x] **Yes** +- [] **No** + +For each task, recovery demonstrations and errors are recorded. + +--- + +## Diversity Dimensions + +- [x] **Target Object** (different wounds for suturing and different anatomical specimens) +- [] **Spatial Layout** +- [x] **Camera Position / Angle** (different position and orientations) +- [x] **Lighting Conditions** +- [ ] **Robot Embodiment** +- [x] **Task Execution** (different operators performed the tasks differently) +- [x] **Background / Scene** (different background conditions) + +We have recorded tasks on a plastic phantom. Camera position and lighting condition were changed after every 20 episodes circa. We recruited 3 expert surgeons and 2 ML researchers that performed the tasks according to their expertise and own techniques. The background conditions change between colon and stomach episodes. + +--- + +## Equipment & Setup + +### Robotic Platform(s) + +- **Robot 1:** dVRK (da Vinci Research Kit) + +### Sensors & Cameras + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | Stereo Endoscope (left), 1920x1080 @ 30 fps | +| **Secondary Camera** | Stereo Endoscope (right), 1920x1080 @ 30 fps | + +--- + +## Action & State Space Representation + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) + +**Action Dimensions:** +``` +action: [psm1_x, psm1_y, psm1_z, psm1_qx, psm1_qy, psm1_qz, psm1_qw, + psm2_x, psm2_y, psm2_z, psm2_qx, psm2_qy, psm2_qz, psm2_qw] +- psm1_x, psm1_y, psm1_z: PSM1 absolute cartesian positions +- psm1_qx, psm1_qy, psm1_qz, psm1_qw: PSM1 absolute cartesian orientations +- psm2_x, psm2_y, psm2_z: PSM2 absolute cartesian positions +- psm2_qx, psm2_qy, psm2_qz, psm2_qw: PSM2 absolute cartesian orientations +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) + +**State Dimensions:** +``` +observation.state: [psm1_j1, psm1_j2, psm1_j3, psm1_j4, psm1_j5, psm1_j6, + psm2_j1, psm2_j2, psm2_j3, psm2_j4, psm2_j5, psm2_j6] +- psm1_j1-j6: PSM1 6-DOF joint positions (radians) +- psm2_j1-j6: PSM2 6-DOF joint positions (radians) +``` + +--- + +## Data Synchronization Approach + +All data are acquired using rosbag and all ROS messages are stamped with their header.stamp fields. +Data is synchronized offline post collection. The right camera serves as the reference timestamp source. For each frame acquired, only the (time) closest data is saved from robot kinematics, with a tolerance of 0.1 secs. + +**Synchronization Method:** +- Reference stream: `/decklink/right/image_raw/compressed` (right stereo camera) +- Tolerance: 0.1 seconds +- Framework: ROS (Robot Operating System) + +Messages from the following topics are synchronized: +- `/decklink/left/image_raw/compressed` - Left camera images +- `/decklink/right/image_raw/compressed` - Right camera images +- `/PSM1/measured_js` - PSM1 joint states +- `/PSM1/measured_cp` - PSM1 absolute cartesian position +- `/PSM2/measured_js` - PSM2 joint states +- `/PSM2/measured_cp` - PSM2 absolute cartesian position + +--- + +## Attribution & Contact + +| | | +| :--- | :--- | +| **Dataset Lead** | Matteo Pescio, Francesco Marzola, Luigi Muratore, Federica Barontini, Giovanni Distefano, Federico Lavagno, Giulio Dagnino, Alberto Arezzo | +| **Institution** | MITIC Lab - UniversitΓ  degli Studi di Torino | +| **Contact Email** | matteo.pescio@unito.it, francesco.marzola@unito.it, luigi.muratore@studenti.polito.it, federica.barontini@unito.it, giovanni.distefano@unito.it, federico.lavagno@unito.it, giulio.dagnino@unito.it, alberto.arezzo@unito.it | +| **Citation (BibTeX)** | 
@misc{dvrk_suturing_subtasks_2025,
  author = {Pescio, Matteo and Marzola, Francesco and Muratore, Luigi and Barontini, Federica and Distefano, Giovanni and Lavagno, Federico and Dagnino, Giulio and Arezzo, Alberto},
  title = {DVRK Suturing Subtasks Dataset},
  year = {2025},
  publisher = {Open-H-Embodiment},
}
| + +--- diff --git a/Surgical/turin/mitic_lerobot_plastic_pad_3dmed/README.md b/Surgical/turin/mitic_lerobot_plastic_pad_3dmed/README.md new file mode 100644 index 0000000000000000000000000000000000000000..3a3de8edea650b2ab89f26436d2c732cb1c9e56a --- /dev/null +++ b/Surgical/turin/mitic_lerobot_plastic_pad_3dmed/README.md @@ -0,0 +1,160 @@ +# DVRK Suturing Subtasks Dataset - README + +--- + +## At a Glance + +Teleoperated demonstrations of a dVRK robot performing suturing subtasksk, like tissue lifting, needle insertion, needle extraction, knot tying on plastic phantom. + +--- + +## Dataset Overview + +| | | +| :--- | :--- | +| **Total Trajectories** | 370 | +| **Total Hours** | [TO be filled] | +| **Data Type** | Ex-Vivo | +| **License** | CC BY 4.0 | +| **Version** | 1.0 | + +--- + +## Tasks & Domain + +### Domain + +- [x] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- Bimanual manipulation +- Pick and place +- Needle handling +- Knot tying +- Soft tissue manipulation + +--- + +## Data Collection Details + +### Collection Method + +- [x] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | 5 operators | +| **Operator Skill Level** | Expert (Surgeons) and Novice (ML researchers with minimal surgical experience) | +| **Collection Period** | [To be filled] | + +### Recovery Demonstrations + +- [x] **Yes** +- [] **No** + +For each task, recovery demonstrations and errors are recorded. + +--- + +## Diversity Dimensions + +- [x] **Target Object** (different wounds for suturing and different anatomical specimens) +- [] **Spatial Layout** +- [x] **Camera Position / Angle** (different position and orientations) +- [x] **Lighting Conditions** +- [ ] **Robot Embodiment** +- [x] **Task Execution** (different operators performed the tasks differently) +- [x] **Background / Scene** (different background conditions) + +We have recorded tasks on a plastic phantom. Camera position and lighting condition were changed after every 20 episodes circa. We recruited 3 expert surgeons and 2 ML researchers that performed the tasks according to their expertise and own techniques. The background conditions change between colon and stomach episodes. + +--- + +## Equipment & Setup + +### Robotic Platform(s) + +- **Robot 1:** dVRK (da Vinci Research Kit) + +### Sensors & Cameras + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | Stereo Endoscope (left), 1920x1080 @ 30 fps | +| **Secondary Camera** | Stereo Endoscope (right), 1920x1080 @ 30 fps | + +--- + +## Action & State Space Representation + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) + +**Action Dimensions:** +``` +action: [psm1_x, psm1_y, psm1_z, psm1_qx, psm1_qy, psm1_qz, psm1_qw, + psm2_x, psm2_y, psm2_z, psm2_qx, psm2_qy, psm2_qz, psm2_qw] +- psm1_x, psm1_y, psm1_z: PSM1 absolute cartesian positions +- psm1_qx, psm1_qy, psm1_qz, psm1_qw: PSM1 absolute cartesian orientations +- psm2_x, psm2_y, psm2_z: PSM2 absolute cartesian positions +- psm2_qx, psm2_qy, psm2_qz, psm2_qw: PSM2 absolute cartesian orientations +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) + +**State Dimensions:** +``` +observation.state: [psm1_j1, psm1_j2, psm1_j3, psm1_j4, psm1_j5, psm1_j6, + psm2_j1, psm2_j2, psm2_j3, psm2_j4, psm2_j5, psm2_j6] +- psm1_j1-j6: PSM1 6-DOF joint positions (radians) +- psm2_j1-j6: PSM2 6-DOF joint positions (radians) +``` + +--- + +## Data Synchronization Approach + +All data are acquired using rosbag and all ROS messages are stamped with their header.stamp fields. +Data is synchronized offline post collection. The right camera serves as the reference timestamp source. For each frame acquired, only the (time) closest data is saved from robot kinematics, with a tolerance of 0.1 secs. + +**Synchronization Method:** +- Reference stream: `/decklink/right/image_raw/compressed` (right stereo camera) +- Tolerance: 0.1 seconds +- Framework: ROS (Robot Operating System) + +Messages from the following topics are synchronized: +- `/decklink/left/image_raw/compressed` - Left camera images +- `/decklink/right/image_raw/compressed` - Right camera images +- `/PSM1/measured_js` - PSM1 joint states +- `/PSM1/measured_cp` - PSM1 absolute cartesian position +- `/PSM2/measured_js` - PSM2 joint states +- `/PSM2/measured_cp` - PSM2 absolute cartesian position + +--- + +## Attribution & Contact + +| | | +| :--- | :--- | +| **Dataset Lead** | Matteo Pescio, Francesco Marzola, Luigi Muratore, Federica Barontini, Giovanni Distefano, Federico Lavagno, Giulio Dagnino, Alberto Arezzo | +| **Institution** | MITIC Lab - UniversitΓ  degli Studi di Torino | +| **Contact Email** | matteo.pescio@unito.it, francesco.marzola@unito.it, luigi.muratore@studenti.polito.it, federica.barontini@unito.it, giovanni.distefano@unito.it, federico.lavagno@unito.it, giulio.dagnino@unito.it, alberto.arezzo@unito.it | +| **Citation (BibTeX)** | 
@misc{dvrk_suturing_subtasks_2025,
  author = {Pescio, Matteo and Marzola, Francesco and Muratore, Luigi and Barontini, Federica and Distefano, Giovanni and Lavagno, Federico and Dagnino, Giulio and Arezzo, Alberto},
  title = {DVRK Suturing Subtasks Dataset},
  year = {2025},
  publisher = {Open-H-Embodiment},
}
| + +--- diff --git a/Surgical/turin/mitic_lerobot_plastic_tube/README.md b/Surgical/turin/mitic_lerobot_plastic_tube/README.md new file mode 100644 index 0000000000000000000000000000000000000000..d024d218854f7dc900e3488c2dda2f9fed4ed6d8 --- /dev/null +++ b/Surgical/turin/mitic_lerobot_plastic_tube/README.md @@ -0,0 +1,160 @@ +# DVRK Suturing Subtasks Dataset - README + +--- + +## At a Glance + +Teleoperated demonstrations of a dVRK robot performing suturing subtasksk, like tissue lifting, needle insertion, needle extraction, knot tying on plastic phantom. + +--- + +## Dataset Overview + +| | | +| :--- | :--- | +| **Total Trajectories** | 480 | +| **Total Hours** | [TO be filled] | +| **Data Type** | Ex-Vivo | +| **License** | CC BY 4.0 | +| **Version** | 1.0 | + +--- + +## Tasks & Domain + +### Domain + +- [x] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- Bimanual manipulation +- Pick and place +- Needle handling +- Knot tying +- Soft tissue manipulation + +--- + +## Data Collection Details + +### Collection Method + +- [x] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | 5 operators | +| **Operator Skill Level** | Expert (Surgeons) and Novice (ML researchers with minimal surgical experience) | +| **Collection Period** | [To be filled] | + +### Recovery Demonstrations + +- [x] **Yes** +- [] **No** + +For each task, recovery demonstrations and errors are recorded. + +--- + +## Diversity Dimensions + +- [x] **Target Object** (different wounds for suturing and different anatomical specimens) +- [] **Spatial Layout** +- [x] **Camera Position / Angle** (different position and orientations) +- [x] **Lighting Conditions** +- [ ] **Robot Embodiment** +- [x] **Task Execution** (different operators performed the tasks differently) +- [x] **Background / Scene** (different background conditions) + +We have recorded tasks on a plastic phantom. Camera position and lighting condition were changed after every 20 episodes circa. We recruited 3 expert surgeons and 2 ML researchers that performed the tasks according to their expertise and own techniques. The background conditions change between colon and stomach episodes. + +--- + +## Equipment & Setup + +### Robotic Platform(s) + +- **Robot 1:** dVRK (da Vinci Research Kit) + +### Sensors & Cameras + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | Stereo Endoscope (left), 1920x1080 @ 30 fps | +| **Secondary Camera** | Stereo Endoscope (right), 1920x1080 @ 30 fps | + +--- + +## Action & State Space Representation + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) + +**Action Dimensions:** +``` +action: [psm1_x, psm1_y, psm1_z, psm1_qx, psm1_qy, psm1_qz, psm1_qw, + psm2_x, psm2_y, psm2_z, psm2_qx, psm2_qy, psm2_qz, psm2_qw] +- psm1_x, psm1_y, psm1_z: PSM1 absolute cartesian positions +- psm1_qx, psm1_qy, psm1_qz, psm1_qw: PSM1 absolute cartesian orientations +- psm2_x, psm2_y, psm2_z: PSM2 absolute cartesian positions +- psm2_qx, psm2_qy, psm2_qz, psm2_qw: PSM2 absolute cartesian orientations +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) + +**State Dimensions:** +``` +observation.state: [psm1_j1, psm1_j2, psm1_j3, psm1_j4, psm1_j5, psm1_j6, + psm2_j1, psm2_j2, psm2_j3, psm2_j4, psm2_j5, psm2_j6] +- psm1_j1-j6: PSM1 6-DOF joint positions (radians) +- psm2_j1-j6: PSM2 6-DOF joint positions (radians) +``` + +--- + +## Data Synchronization Approach + +All data are acquired using rosbag and all ROS messages are stamped with their header.stamp fields. +Data is synchronized offline post collection. The right camera serves as the reference timestamp source. For each frame acquired, only the (time) closest data is saved from robot kinematics, with a tolerance of 0.1 secs. + +**Synchronization Method:** +- Reference stream: `/decklink/right/image_raw/compressed` (right stereo camera) +- Tolerance: 0.1 seconds +- Framework: ROS (Robot Operating System) + +Messages from the following topics are synchronized: +- `/decklink/left/image_raw/compressed` - Left camera images +- `/decklink/right/image_raw/compressed` - Right camera images +- `/PSM1/measured_js` - PSM1 joint states +- `/PSM1/measured_cp` - PSM1 absolute cartesian position +- `/PSM2/measured_js` - PSM2 joint states +- `/PSM2/measured_cp` - PSM2 absolute cartesian position + +--- + +## Attribution & Contact + +| | | +| :--- | :--- | +| **Dataset Lead** | Matteo Pescio, Francesco Marzola, Luigi Muratore, Federica Barontini, Giovanni Distefano, Federico Lavagno, Giulio Dagnino, Alberto Arezzo | +| **Institution** | MITIC Lab - UniversitΓ  degli Studi di Torino | +| **Contact Email** | matteo.pescio@unito.it, francesco.marzola@unito.it, luigi.muratore@studenti.polito.it, federica.barontini@unito.it, giovanni.distefano@unito.it, federico.lavagno@unito.it, giulio.dagnino@unito.it, alberto.arezzo@unito.it | +| **Citation (BibTeX)** | 
@misc{dvrk_suturing_subtasks_2025,
  author = {Pescio, Matteo and Marzola, Francesco and Muratore, Luigi and Barontini, Federica and Distefano, Giovanni and Lavagno, Federico and Dagnino, Giulio and Arezzo, Alberto},
  title = {DVRK Suturing Subtasks Dataset},
  year = {2025},
  publisher = {Open-H-Embodiment},
}
| + +--- diff --git a/Surgical/ucsd/surgical_learning_dataset/README.md b/Surgical/ucsd/surgical_learning_dataset/README.md new file mode 100644 index 0000000000000000000000000000000000000000..f532466773ae7313c8cc23cd5c4e9066d9477605 --- /dev/null +++ b/Surgical/ucsd/surgical_learning_dataset/README.md @@ -0,0 +1,227 @@ +# Retraction and Dissection Dataset + +--- + +## πŸ“‹ At a Glance + +Teleoperated demonstrations of surgical retraction and tissue dissection tasks collected for learning-based robotic manipulation. + +--- + +## πŸ“– Dataset Overview + +This dataset contains expert teleoperated demonstrations of **retraction** and **dissection**, two fundamental subtasks in surgical manipulation workflows. +The dataset is designed to support **imitation learning**, **policy learning**, and **multi-skill robotic manipulation** in contact-rich and constrained environments. + +Demonstrations include both successful executions and recovery behaviors from suboptimal states, enabling robust policy learning under realistic failure conditions. + +| | | +| :--- | :--- | +| **Total Trajectories** | 1200 | +| **Total Hours** | 18.5 | +| **Data Type** | [ ] Clinical [ ] Ex-Vivo [x] Table-Top Phantom [ ] Digital Simulation [ ] Physical Simulation | +| **License** | CC BY 4.0 | +| **Version** | 1.0 | + +--- + +## 🎯 Tasks & Domain + +### Domain + +- [x] Surgical Robotics +- [ ] Ultrasound Robotics +- [ ] Other Healthcare Robotics + +### Demonstrated Skills + +This dataset focuses on two complementary surgical manipulation skills: + +- **Retraction** + - Grasping deformable tissue or surrogate material + - Maintaining stable and continuous tension + - Adjusting pulling direction to expose the target region + +- **Dissection** + - Tool alignment and approach + - Controlled cutting along tissue boundaries + - Coordinated motion between retraction and cutting tools + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +- [x] Human Teleoperation +- [ ] Programmatic / State-Machine +- [ ] AI Policy / Autonomous + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | 2 | +| **Operator Skill Level** | Intermediate (Trained Researcher) | +| **Collection Period** | 2024-06-01 to 2024-08-30 | + +### Recovery Demonstrations + +- [x] Yes +- [ ] No + +**Description:** +Recovery demonstrations include failed grasps, insufficient tissue tension, and misaligned cutting trajectories. +Operators re-orient tools, re-establish stable contact, and resume task execution from intermediate states. + +--- + +## πŸ’‘ Diversity Dimensions + +- [x] Camera Position / Angle +- [ ] Lighting Conditions +- [x] Target Object +- [x] Spatial Layout +- [ ] Robot Embodiment +- [x] Task Execution +- [ ] Background / Scene + +**Elaboration:** +Target tissue phantoms are placed at varying positions and orientations. +Operators employ different retraction directions and dissection strategies to encourage behavioral diversity. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform + +- da Vinci Research Kit (dVRK) + +### Sensors & Cameras + +| Type | Model / Details | +| :--- | :--- | +| Primary Camera | Endoscopic camera, 1920Γ—1080 @ 30fps | +| Room Camera | Fixed external RGB camera | +| Force/Torque Sensor | Not available | +| Other Sensors | Robot joint encoders | + +--- + +## 🎯 Action & State Space Representation + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [x] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** +- [ ] **Axis-Angle** +- [ ] **Rotation Matrix** +- [ ] **Other** + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [x] **Other** (Delta end-effector pose commands; reference frame not explicitly specified) + +**Action Dimensions:** +``` +action: [ + dPSM_RETRACTION_x, + dPSM_RETRACTION_y, + dPSM_RETRACTION_z, + dPSM_RETRACTION_qw, + dPSM_RETRACTION_qx, + dPSM_RETRACTION_qy, + dPSM_RETRACTION_qz, + dPSM_RETRACTION_gripper, + dPSM_CUTTER_x, + dPSM_CUTTER_y, + dPSM_CUTTER_z, + dPSM_CUTTER_qw, + dPSM_CUTTER_qx, + dPSM_CUTTER_qy, + dPSM_CUTTER_qz, + dPSM_CUTTER_gripper +] +``` + +- 16-D float32 action vector per timestep +- Relative end-effector pose updates for two PSMs +- Quaternion-based orientation deltas +- Gripper control commands + +--- + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** +- [x] **Joint Velocities** +- [x] **End-Effector Pose** +- [ ] **Force/Torque Readings** +- [x] **Gripper State** +- [x] **Other** (joint efforts, dissection target points) + +**State Dimensions:** +``` +observation.state: [ + psm_retraction_joint_positions (6), + psm_retraction_joint_velocities (6), + psm_retraction_joint_efforts (6), + psm_retraction_gripper_state (2), + psm_cutter_joint_positions (6), + psm_cutter_joint_velocities (6), + psm_cutter_joint_efforts (6), + psm_cutter_gripper_state (2), + retraction_end_effector_pose (7), + cutter_end_effector_pose (7), + dissection_target_points (8) +] +``` + +- 62-D float32 state vector +- Full kinematic + effort feedback for both PSMs +- End-effector poses as position + quaternion +- Four 2D dissection target points + +## ⏱️ Data Synchronization Approach + +All robot state signals, action commands, and stereo endoscopic video streams are synchronized using a shared system clock during data collection. +Each frame is timestamped at capture time and aligned at a fixed rate of **30 FPS**. + +Specifically: +- Robot joint states, end-effector poses, gripper states, and action commands are recorded at the control loop frequency. +- Stereo endoscopic videos (`observation.images.left` and `observation.images.right`) are captured at 30 FPS and timestamped using the same clock. +- During dataset export, all modalities are aligned by timestamp to form per-frame observations stored in parquet files. + +This ensures consistent temporal correspondence between visual observations, robot kinematics, and control actions across the entire dataset. + +--- + +## πŸ‘₯ Attribution & Contact + +| | | +| :--- | :--- | +| **Dataset Lead** | Changwei Chen, Yinuo Yang, Xiao Liang, Michael Yip | +| **Institution** | Department of Electrical and Computer Engineering, University of California San Diego, La Jolla, CA 92093, USA | +| **Contact Email** | {chc165, yiy124 ,x5liang, yip}@ucsd.edu | +| **Citation (BibTeX)** | @misc{ucsd_openh_phantom_2026, author = {Chen, Changwei and Yang, Yinuo and Liang, Xiao and Yip, Michael}, title = {UC San Diego Phantom Surgical Robotics Dataset (Open-H Embodiment)}, year = {2026}, publisher = {Open-H Embodiment}, note = {Dataset will be released via Open-H Embodiment; link/DOI forthcoming. Contact: chc165@ucsd.edu} +} |## πŸ‘₯ Attribution & Contact + +| | | +| :--- | :--- | +| **Dataset Lead** | Changwei Chen, Yinuo Yang, Xiao Liang, Michael Yip | +| **Institution** | Department of Electrical and Computer Engineering, University of California San Diego, La Jolla, CA 92093, USA | +| **Contact Email** | {chc165, yiy124 ,x5liang, yip}@ucsd.edu | +| **Citation (BibTeX)** | @misc{ucsd_openh_phantom_2026, author = {Chen, Changwei and Yang, Yinuo and Liang, Xiao and Yip, Michael}, title = {UC San Diego Phantom Surgical Robotics Dataset (Open-H Embodiment)}, year = {2026}, publisher = {Open-H Embodiment}, note = {Dataset will be released via Open-H Embodiment; link/DOI forthcoming. Contact: chc165@ucsd.edu} +} | diff --git a/Surgical/ucsd/surgical_learning_dataset2/README.md b/Surgical/ucsd/surgical_learning_dataset2/README.md new file mode 100644 index 0000000000000000000000000000000000000000..71cae6cb5ccddfc8bb2b06fd7c3c1f27c731884f --- /dev/null +++ b/Surgical/ucsd/surgical_learning_dataset2/README.md @@ -0,0 +1,211 @@ +# Retraction Dataset + +--- + +## πŸ“‹ At a Glance + +Teleoperated demonstrations of surgical retraction tasks collected for learning-based robotic manipulation on the da Vinci Research Kit (dVRK). + +--- + +## πŸ“– Dataset Overview + +This dataset contains expert teleoperated demonstrations of **retraction**, a fundamental subtask in surgical manipulation workflows. +The dataset is designed to support **imitation learning**, **policy learning**, and **robot state–conditioned control** in contact-rich and constrained environments. + +Demonstrations focus on stable tissue grasping and tension adjustment behaviors. + +| | | +| :--- | :--- | +| **Total Trajectories** | 200 | +| **Total Hours** | ~0.24 | +| **Data Type** | [ ] Clinical [ ] Ex-Vivo [x] Table-Top Phantom [ ] Digital Simulation [ ] Physical Simulation | +| **Robot Platform** | da Vinci Research Kit (dVRK) | +| **FPS** | 30 | +| **License** | CC BY 4.0 | +| **Version** | 1.0 | + +--- + +## 🎯 Tasks & Domain + +### Domain + +- [x] Surgical Robotics +- [ ] Ultrasound Robotics +- [ ] Other Healthcare Robotics + +### Demonstrated Skills + +This dataset focuses on a single surgical manipulation skill: + +- **Retraction** + - Grasping deformable tissue or surrogate material + - Maintaining stable and continuous tension + - Adjusting pulling direction to expose target regions + +Task semantics are provided via `instruction.text` for each episode. + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +- [x] Human Teleoperation +- [ ] Programmatic / State-Machine +- [ ] AI Policy / Autonomous + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | 2 | +| **Operator Skill Level** | Intermediate (Trained Researcher) | +| **Collection Period** | 2024-06-01 to 2024-08-30 | + +### Recovery Demonstrations + +- [ ] Yes +- [x] No + +--- + +## πŸ’‘ Diversity Dimensions + +- [ ] Camera Position / Angle +- [ ] Lighting Conditions +- [x] Target Object +- [x] Spatial Layout +- [ ] Robot Embodiment +- [x] Task Execution +- [ ] Background / Scene + +**Elaboration:** +Target tissue phantoms are placed at varying positions and orientations. +Operators employ different retraction directions to encourage behavioral diversity. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform + +- **da Vinci Research Kit (dVRK)** + - Dual PSM configuration: + - Bipolar Forceps (Retraction) + - Potts Scissors (present but not actively used for cutting in this dataset) + +### Sensors & Cameras + +| Type | Model / Details | +| :--- | :--- | +| Primary Camera | Stereo endoscopic cameras, 480Γ—640 @ 30fps | +| Room Camera | Not used | +| Force/Torque Sensor | Not available | +| Other Sensors | Robot joint encoders | + +--- + +## 🎯 Action & State Space Representation + +### Action Space Representation + +**Primary Action Representation:** +- [ ] Absolute Cartesian +- [x] Relative Cartesian (delta end-effector pose) +- [ ] Joint Space +- [ ] Other + +**Orientation Representation:** +- [x] Quaternions (qw, qx, qy, qz) +- [ ] **Euler Angles** +- [ ] **Axis-Angle** +- [ ] **Rotation Matrix** +- [ ] **Other** + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [x] Other (delta end-effector pose commands; reference frame not explicitly specified) + +**Action Dimensions:** +``` +action: [ + dPSM_RETRACTION_x, + dPSM_RETRACTION_y, + dPSM_RETRACTION_z, + dPSM_RETRACTION_qw, + dPSM_RETRACTION_qx, + dPSM_RETRACTION_qy, + dPSM_RETRACTION_qz, + dPSM_RETRACTION_gripper, + dPSM_CUTTER_x, + dPSM_CUTTER_y, + dPSM_CUTTER_z, + dPSM_CUTTER_qw, + dPSM_CUTTER_qx, + dPSM_CUTTER_qy, + dPSM_CUTTER_qz, + dPSM_CUTTER_gripper +] +``` + +- 16-D float32 action vector per timestep +- Relative end-effector pose updates for two PSMs +- Quaternion-based orientation deltas +- Gripper control commands + +--- + +### State Space Representation + +**State Information Included:** +- [x] Joint Positions +- [x] End-Effector Pose +- [x] Gripper State +- [ ] Joint Velocities +- [ ] Joint Efforts +- [ ] Force/Torque Readings + +**State Dimensions (28D):** +``` +observation.state: [ + psm_retraction_joint_positions (6), + psm_retraction_gripper_pos (1), + psm_cutter_joint_positions (6), + psm_cutter_gripper_pos (1), + retraction_end_effector_pose (7), + cutter_end_effector_pose (7) +] +``` + +- `observation.state` is a 28-D float32 vector +- Contains joint positions, gripper states, and EE poses +- No velocities or effort terms are included + +--- + +## ⏱️ Data Synchronization Approach + +All robot state signals, action commands, and stereo endoscopic video streams are synchronized using a shared system clock during data collection. + +- Robot states and actions are recorded at the control frequency. +- Stereo endoscopic videos are captured at **30 FPS**. +- During dataset export, all modalities are aligned by timestamp and stored per frame in parquet format. + +This ensures consistent temporal correspondence between visual observations and robot kinematics across the dataset. + +--- + +## πŸ‘₯ Attribution & Contact + +| | | +| :--- | :--- | +| **Dataset Lead** | Changwei Chen, Yinuo Yang, Xiao Liang, Michael Yip | +| **Institution** | Department of Electrical and Computer Engineering, University of California San Diego, La Jolla, CA 92093, USA | +| **Contact Email** | {chc165, yiy124 ,x5liang, yip}@ucsd.edu | +| **Citation (BibTeX)** | @misc{ucsd_openh_phantom_2026, author = {Chen, Changwei and Yang, Yinuo and Liang, Xiao and Yip, Michael}, title = {UC San Diego Phantom Surgical Robotics Dataset (Open-H Embodiment)}, year = {2026}, publisher = {Open-H Embodiment}, note = {Dataset will be released via Open-H Embodiment; link/DOI forthcoming. Contact: chc165@ucsd.edu} +} | diff --git a/Surgical/ucsd/surgical_learning_retraction_dataset3/README.md b/Surgical/ucsd/surgical_learning_retraction_dataset3/README.md new file mode 100644 index 0000000000000000000000000000000000000000..e9a6f7d11fef593c65b4bc2a629828315987e22d --- /dev/null +++ b/Surgical/ucsd/surgical_learning_retraction_dataset3/README.md @@ -0,0 +1,215 @@ +# Retraction and Dissection Dataset + +--- + +## πŸ“‹ At a Glance + +Teleoperated demonstrations of surgical retraction tasks collected for learning-based robotic manipulation on the da Vinci Research Kit (dVRK). + +--- + +## πŸ“– Dataset Overview + +This dataset contains expert teleoperated demonstrations of **retraction**, a fundamental subtask in surgical manipulation workflows. +The dataset is designed to support **imitation learning**, **policy learning**, and **robot state–conditioned control** in contact-rich and constrained environments. + +Demonstrations focus on stable tissue grasping and tension adjustment behaviors. + +| | | +| :--- | :--- | +| **Total Trajectories** | 598 | +| **Total Hours** | ~1.70 | +| **Data Type** | [ ] Clinical [ ] Ex-Vivo [x] Table-Top Phantom [ ] Digital Simulation [ ] Physical Simulation | +| **Robot Platform** | da Vinci Research Kit (dVRK) | +| **FPS** | 30 | +| **License** | CC BY 4.0 | +| **Version** | 1.0 | + +--- + +## 🎯 Tasks & Domain + +### Domain + +- [x] Surgical Robotics +- [ ] Ultrasound Robotics +- [ ] Other Healthcare Robotics + +### Demonstrated Skills + +This dataset focuses on a single surgical manipulation skill: + +- **Retraction** + - Grasping deformable tissue or surrogate material + - Maintaining stable and continuous tension + - Adjusting pulling direction to expose target regions + +Task semantics are provided via `instruction.text` for each episode. + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +- [x] Human Teleoperation +- [ ] Programmatic / State-Machine +- [ ] AI Policy / Autonomous + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | 2 | +| **Operator Skill Level** | Intermediate (Trained Researcher) | +| **Collection Period** | 2024-06-01 to 2024-08-30 | + +### Recovery Demonstrations + +- [ ] Yes +- [x] No + +--- + +## πŸ’‘ Diversity Dimensions + +- [ ] Camera Position / Angle +- [ ] Lighting Conditions +- [x] Target Object +- [x] Spatial Layout +- [ ] Robot Embodiment +- [x] Task Execution +- [ ] Background / Scene + +**Elaboration:** +Target tissue phantoms are placed at varying positions and orientations. +Operators employ different retraction directions to encourage behavioral diversity. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform + +- **da Vinci Research Kit (dVRK)** + - Dual PSM configuration: + - Bipolar Forceps (Retraction) + - Potts Scissors (present but not actively used for cutting in this dataset) + +### Sensors & Cameras + +| Type | Model / Details | +| :--- | :--- | +| Primary Camera | Stereo endoscopic cameras, 480Γ—640 @ 30fps | +| Room Camera | Not used | +| Force/Torque Sensor | Not available | +| Other Sensors | Robot joint encoders | + +--- + +## 🎯 Action & State Space Representation + +### Action Space Representation + +**Primary Action Representation:** +- [ ] Absolute Cartesian +- [x] Relative Cartesian (delta end-effector pose) +- [ ] Joint Space +- [ ] Other + +**Orientation Representation:** +- [x] Quaternions (qw, qx, qy, qz) +- [ ] **Euler Angles** +- [ ] **Axis-Angle** +- [ ] **Rotation Matrix** +- [ ] **Other** + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [x] Other (delta end-effector pose commands; reference frame not explicitly specified) + +**Action Dimensions (16D):** +``` +action: [ + dPSM_RETRACTION_x, + dPSM_RETRACTION_y, + dPSM_RETRACTION_z, + dPSM_RETRACTION_qw, + dPSM_RETRACTION_qx, + dPSM_RETRACTION_qy, + dPSM_RETRACTION_qz, + dPSM_RETRACTION_gripper, + dPSM_CUTTER_x, + dPSM_CUTTER_y, + dPSM_CUTTER_z, + dPSM_CUTTER_qw, + dPSM_CUTTER_qx, + dPSM_CUTTER_qy, + dPSM_CUTTER_qz, + dPSM_CUTTER_gripper +] +``` + +- 16-D float32 action vector per timestep +- Relative end-effector pose updates for two PSMs +- Quaternion-based orientation deltas +- Gripper control commands + +--- + +### State Space Representation + +**State Information Included:** +- [x] Joint Positions +- [x] Joint Velocities +- [x] Joint Efforts +- [x] End-Effector Pose +- [x] Gripper State +- [ ] Force/Torque Readings + +**State Dimensions (54D):** +``` +observation.state: [ + psm_retraction_joint_positions (6), + psm_retraction_joint_velocities (6), + psm_retraction_joint_efforts (6), + psm_retraction_gripper_state (2), + psm_cutter_joint_positions (6), + psm_cutter_joint_velocities (6), + psm_cutter_joint_efforts (6), + psm_cutter_gripper_state (2), + retraction_end_effector_pose (7), + cutter_end_effector_pose (7) +] +``` + +- `observation.state` is a 54-D float32 vector +- Includes joint positions, velocities, efforts, gripper states, and EE poses +- No external force/torque sensors are used + +--- + +## ⏱️ Data Synchronization Approach + +All robot state signals, action commands, and stereo endoscopic video streams are synchronized using a shared system clock during data collection. + +- Robot states and actions are recorded at the control frequency. +- Stereo endoscopic videos are captured at **30 FPS**. +- During dataset export, all modalities are aligned by timestamp and stored per frame in parquet format. + +This ensures consistent temporal correspondence between visual observations and robot kinematics across the dataset. + +--- + +## πŸ‘₯ Attribution & Contact + +| | | +| :--- | :--- | +| **Dataset Lead** | Changwei Chen, Yinuo Yang, Xiao Liang, Michael Yip | +| **Institution** | Department of Electrical and Computer Engineering, University of California San Diego, La Jolla, CA 92093, USA | +| **Contact Email** | {chc165, yiy124 ,x5liang, yip}@ucsd.edu | +| **Citation (BibTeX)** | @misc{ucsd_openh_phantom_2026, author = {Chen, Changwei and Yang, Yinuo and Liang, Xiao and Yip, Michael}, title = {UC San Diego Phantom Surgical Robotics Dataset (Open-H Embodiment)}, year = {2026}, publisher = {Open-H Embodiment}, note = {Dataset will be released via Open-H Embodiment; link/DOI forthcoming. Contact: chc165@ucsd.edu} +} | diff --git a/Surgical/ucsd/surgical_learning_retraction_failurecase/README.md b/Surgical/ucsd/surgical_learning_retraction_failurecase/README.md new file mode 100644 index 0000000000000000000000000000000000000000..37fa999a9fde82e0e5d7ff22f7381e664ddf809f --- /dev/null +++ b/Surgical/ucsd/surgical_learning_retraction_failurecase/README.md @@ -0,0 +1,215 @@ +# Retraction Failure-Case Dataset + +--- + +## πŸ“‹ At a Glance + +Teleoperated **failure-case** demonstrations of surgical retraction collected for learning-based robotic manipulation on the da Vinci Research Kit (dVRK). + +--- + +## πŸ“– Dataset Overview + +This dataset contains expert teleoperated demonstrations of **retraction failure cases**, capturing unsuccessful or suboptimal executions during surgical manipulation workflows. +The dataset is designed to support research in **robust imitation learning**, **failure recovery**, **error-aware policy learning**, and **robot state–conditioned control** in contact-rich and constrained environments. + +Demonstrations include loss of stable grasp, insufficient tissue tension, slippage, and misaligned pulling directions, without successful task completion. + +| | | +| :--- | :--- | +| **Total Trajectories** | 299 | +| **Total Hours** | ~0.62 | +| **Data Type** | [ ] Clinical [ ] Ex-Vivo [x] Table-Top Phantom [ ] Digital Simulation [ ] Physical Simulation | +| **Robot Platform** | da Vinci Research Kit (dVRK) | +| **FPS** | 30 | +| **License** | CC BY 4.0 | +| **Version** | 1.0 | + +--- + +## 🎯 Tasks & Domain + +### Domain + +- [x] Surgical Robotics +- [ ] Ultrasound Robotics +- [ ] Other Healthcare Robotics + +### Demonstrated Skills + +This dataset focuses on **failure cases of a single surgical manipulation skill**: + +- **Retraction (Failure Cases)** + - Unstable or failed tissue grasp + - Insufficient or inconsistent tension application + - Slippage during pulling + - Misaligned pulling direction leading to task failure + +Task semantics are provided via `instruction.text` for each episode and indicate failure or unsuccessful execution. + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +- [x] Human Teleoperation +- [ ] Programmatic / State-Machine +- [ ] AI Policy / Autonomous + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | 2 | +| **Operator Skill Level** | Intermediate (Trained Researcher) | +| **Collection Period** | 2024-06-01 to 2024-08-30 | + +### Recovery Demonstrations + +- [ ] Yes +- [x] No + +*Note: Episodes terminate upon task failure and do not include corrective recovery behaviors.* + +--- + +## πŸ’‘ Diversity Dimensions + +- [ ] Camera Position / Angle +- [ ] Lighting Conditions +- [x] Target Object +- [x] Spatial Layout +- [ ] Robot Embodiment +- [x] Failure Mode +- [ ] Background / Scene + +**Elaboration:** +Failures arise from diverse initial conditions, tissue placements, and operator-induced perturbations, resulting in varied failure modes across episodes. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform + +- **da Vinci Research Kit (dVRK)** + - Dual PSM configuration: + - Bipolar Forceps (Retraction) + - Potts Scissors (present but not actively used) + +### Sensors & Cameras + +| Type | Model / Details | +| :--- | :--- | +| Primary Camera | Stereo endoscopic cameras, 480Γ—640 @ 30fps | +| Room Camera | Not used | +| Force/Torque Sensor | Not available | +| Other Sensors | Robot joint encoders | + +--- + +## 🎯 Action & State Space Representation + +### Action Space Representation + +**Primary Action Representation:** +- [ ] Absolute Cartesian +- [x] Relative Cartesian (delta end-effector pose) +- [ ] Joint Space +- [ ] Other + +**Orientation Representation:** +- [x] Quaternions (qw, qx, qy, qz) +- [ ] **Euler Angles** +- [ ] **Axis-Angle** +- [ ] **Rotation Matrix** +- [ ] **Other** + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [x] Other (delta end-effector pose commands; reference frame not explicitly specified) + +**Action Dimensions (16D):** +``` +action: [ + dPSM_RETRACTION_x, + dPSM_RETRACTION_y, + dPSM_RETRACTION_z, + dPSM_RETRACTION_qw, + dPSM_RETRACTION_qx, + dPSM_RETRACTION_qy, + dPSM_RETRACTION_qz, + dPSM_RETRACTION_gripper, + dPSM_CUTTER_x, + dPSM_CUTTER_y, + dPSM_CUTTER_z, + dPSM_CUTTER_qw, + dPSM_CUTTER_qx, + dPSM_CUTTER_qy, + dPSM_CUTTER_qz, + dPSM_CUTTER_gripper +] +``` + +- Relative end-effector pose updates per timestep +- Quaternion-based orientation deltas +- Gripper control commands + +--- + +### State Space Representation + +**State Information Included:** +- [x] Joint Positions +- [x] Joint Velocities +- [x] Joint Efforts +- [x] End-Effector Pose +- [x] Gripper State +- [ ] Force/Torque Readings + +**State Dimensions (54D):** +``` +observation.state: [ + psm_retraction_joint_positions (6), + psm_retraction_joint_velocities (6), + psm_retraction_joint_efforts (6), + psm_retraction_gripper_state (2), + psm_cutter_joint_positions (6), + psm_cutter_joint_velocities (6), + psm_cutter_joint_efforts (6), + psm_cutter_gripper_state (2), + retraction_end_effector_pose (7), + cutter_end_effector_pose (7) +] +``` + +- `observation.state` is a 54-D float32 vector +- Includes joint positions, velocities, efforts, gripper states, and EE poses + +--- + +## ⏱️ Data Synchronization Approach + +All robot state signals, action commands, and stereo endoscopic video streams are synchronized using a shared system clock during data collection. + +- Robot states and actions are recorded at the control frequency. +- Stereo endoscopic videos are captured at **30 FPS**. +- During dataset export, all modalities are aligned by timestamp and stored per frame in parquet format. + +This ensures consistent temporal correspondence between visual observations and robot kinematics across failure-case episodes. + +--- + +## πŸ‘₯ Attribution & Contact + +| | | +| :--- | :--- | +| **Dataset Lead** | Changwei Chen, Yinuo Yang, Xiao Liang, Michael Yip | +| **Institution** | Department of Electrical and Computer Engineering, University of California San Diego, La Jolla, CA 92093, USA | +| **Contact Email** | {chc165, yiy124 ,x5liang, yip}@ucsd.edu | +| **Citation (BibTeX)** | @misc{ucsd_openh_phantom_2026, author = {Chen, Changwei and Yang, Yinuo and Liang, Xiao and Yip, Michael}, title = {UC San Diego Phantom Surgical Robotics Dataset (Open-H Embodiment)}, year = {2026}, publisher = {Open-H Embodiment}, note = {Dataset will be released via Open-H Embodiment; link/DOI forthcoming. Contact: chc165@ucsd.edu} +} | diff --git a/Ultrasound/balgrist/sonogym_open_h_us_guidance_l1/README.md b/Ultrasound/balgrist/sonogym_open_h_us_guidance_l1/README.md new file mode 100644 index 0000000000000000000000000000000000000000..9eae8c1615ff5c8b9c9426e3890b2c8d4809fe7e --- /dev/null +++ b/Ultrasound/balgrist/sonogym_open_h_us_guidance_l1/README.md @@ -0,0 +1,231 @@ + + +# SonoGym Probe Manipulation Lerobot Dataset_1 - README + +--- + +## πŸ“‹ At a Glance + +*Synthetic ultrasound probe manipulation to see L1 vertebra.* + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 1024 trajectories of expert policies to move the probe to be above the target vertebra.* + +| | | +| :--- | :--- | +| **Total Trajectories** | `[1024]` | +| **Total Hours** | `[]` | +| **Data Type** | `[ ] Clinical` `[] Ex-Vivo` `[ ] Table-Top Phantom` `[x] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +- Navigate ultrasound prove to find a specific anatomy + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [x] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[x] N/A` | +| **Collection Period** | From `[2025-03]` to `[2026-01]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + +**If yes, please briefly describe the recovery process:** + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [x] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +We start the ultrasound scan from various different initial positions. We scan multiple bone structures including vertebra L1-L5. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +*Kuka med14* + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `[None]` | +| **Room/3rd Person Camera** | `[None]` | +| **Force/Torque Sensor** | `[None]` | +| **Medical Imager** | `[e.g., Synthetic ultrasound by GAN, B-Mode]` | +| **Other** | `[Specify]` | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [x] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `None`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [x] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* +``` +action: [x, y, z, qw, qx, qy, qz] +- x, y, z: relative position in ultrasound image frame (meters) to next pose. +- qw, qx, qy, qz: quaternion rotation to the next frame +- Note that this 3D motion is along the surface of the training patient model. This may not align with the testing case. +``` + + +**Example:** +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +``` +observation.state.ee_state: [x, y, z, qw, qx, qy, qz] +- x, y, z: Absolute position in base frame (meters) +- qw, qx, qy, qz: quaternion in base frame +observation.state.joint_state: [j1, j2, j3, j4, j5, j6, j7] +- Absolute joint positions for 7-DOF arm (radians) +``` + + +**Example:** +``` +observation.state: [j1, j2, j3, j4, j5, j6, j7, gripper_pos] +- j1-j7: Absolute joint positions for 7-DOF arm (radians) +- gripper_pos: Current gripper opening (meters) +``` + +### πŸ“‹ Recommended Additional Representations + +*Even if not your primary action/state representation, we strongly encourage including these standardized formats for maximum compatibility:* + +**Recommended Action Fields:** +- **`action.cartesian_absolute`**: Absolute Cartesian pose with absolute quaternions + ``` + [x, y, z, qx, qy, qz, qw, gripper_angle] + ``` + +**Recommended State Fields:** +- **`observation.state.joint_positions`**: Absolute positions for all articulated joints + ``` + [joint_1, joint_2, ..., joint_n] + ``` + + +--- + +## ⏱️ Data Synchronization Approach + + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +*The synchronization is ensured automatically through simulation*. + +**Example:** *We collect joint kinematics from our Franka Research 3 and RGB-D frames from Intel RealSense D435 cameras, all running in ROS 2 Galactic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /camera/*/image_raw, and /camera/*/camera_info in a single rosbag2 session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±2 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Yunke Ao, Luohong Wu]` | +| **Institution** | `[Balgrist University Hospital]` | +| **Contact Email** | `[yunke.ao@balgrist.ch, luohong.wu@balgrist.ch, ...]` | +| **Citation (BibTeX)** | 
@misc{[Ultrabones100k_lerobot_2026],
  author = {[Yunke Ao, Luohong Wu, Philipp Fuernstahl]},
  title = {[Ultrabones100k Lerobot Dataset]},
  year = {2026},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Ultrasound/balgrist/sonogym_open_h_us_guidance_l2/README.md b/Ultrasound/balgrist/sonogym_open_h_us_guidance_l2/README.md new file mode 100644 index 0000000000000000000000000000000000000000..a0325934d12e13a0e46453d4eb9bd519c33a5323 --- /dev/null +++ b/Ultrasound/balgrist/sonogym_open_h_us_guidance_l2/README.md @@ -0,0 +1,231 @@ + + +# SonoGym Probe Manipulation Lerobot Dataset_2 - README + +--- + +## πŸ“‹ At a Glance + +*Synthetic ultrasound probe manipulation to see L2 vertebra.* + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 1024 trajectories of expert policies to move the probe to be above the target vertebra.* + +| | | +| :--- | :--- | +| **Total Trajectories** | `[1024]` | +| **Total Hours** | `[]` | +| **Data Type** | `[ ] Clinical` `[] Ex-Vivo` `[ ] Table-Top Phantom` `[x] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +- Navigate ultrasound prove to find a specific anatomy + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [x] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[x] N/A` | +| **Collection Period** | From `[2025-03]` to `[2026-01]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + +**If yes, please briefly describe the recovery process:** + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [x] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +We start the ultrasound scan from various different initial positions. We scan multiple bone structures including vertebra L1-L5. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +*Kuka med14* + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `[None]` | +| **Room/3rd Person Camera** | `[None]` | +| **Force/Torque Sensor** | `[None]` | +| **Medical Imager** | `[e.g., Synthetic ultrasound by GAN, B-Mode]` | +| **Other** | `[Specify]` | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [x] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `None`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [x] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* +``` +action: [x, y, z, qw, qx, qy, qz] +- x, y, z: relative position in ultrasound image frame (meters) to next pose. +- qw, qx, qy, qz: quaternion rotation to the next frame +- Note that this 3D motion is along the surface of the training patient model. This may not align with the testing case. +``` + + +**Example:** +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +``` +observation.state.ee_state: [x, y, z, qw, qx, qy, qz] +- x, y, z: Absolute position in base frame (meters) +- qw, qx, qy, qz: quaternion in base frame +observation.state.joint_state: [j1, j2, j3, j4, j5, j6, j7] +- Absolute joint positions for 7-DOF arm (radians) +``` + + +**Example:** +``` +observation.state: [j1, j2, j3, j4, j5, j6, j7, gripper_pos] +- j1-j7: Absolute joint positions for 7-DOF arm (radians) +- gripper_pos: Current gripper opening (meters) +``` + +### πŸ“‹ Recommended Additional Representations + +*Even if not your primary action/state representation, we strongly encourage including these standardized formats for maximum compatibility:* + +**Recommended Action Fields:** +- **`action.cartesian_absolute`**: Absolute Cartesian pose with absolute quaternions + ``` + [x, y, z, qx, qy, qz, qw, gripper_angle] + ``` + +**Recommended State Fields:** +- **`observation.state.joint_positions`**: Absolute positions for all articulated joints + ``` + [joint_1, joint_2, ..., joint_n] + ``` + + +--- + +## ⏱️ Data Synchronization Approach + + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +*The synchronization is ensured automatically through simulation*. + +**Example:** *We collect joint kinematics from our Franka Research 3 and RGB-D frames from Intel RealSense D435 cameras, all running in ROS 2 Galactic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /camera/*/image_raw, and /camera/*/camera_info in a single rosbag2 session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±2 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Yunke Ao, Luohong Wu]` | +| **Institution** | `[Balgrist University Hospital]` | +| **Contact Email** | `[yunke.ao@balgrist.ch, luohong.wu@balgrist.ch, ...]` | +| **Citation (BibTeX)** | 
@misc{[Ultrabones100k_lerobot_2026],
  author = {[Yunke Ao, Luohong Wu, Philipp Fuernstahl]},
  title = {[Ultrabones100k Lerobot Dataset]},
  year = {2026},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Ultrasound/balgrist/sonogym_open_h_us_guidance_l3/README.md b/Ultrasound/balgrist/sonogym_open_h_us_guidance_l3/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b9cd0f0b36f71885c6f0c29bdedc28e08dbcbffd --- /dev/null +++ b/Ultrasound/balgrist/sonogym_open_h_us_guidance_l3/README.md @@ -0,0 +1,231 @@ + + +# SonoGym Probe Manipulation Lerobot Dataset_3 - README + +--- + +## πŸ“‹ At a Glance + +*Synthetic ultrasound probe manipulation to see L3 vertebra.* + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 1024 trajectories of expert policies to move the probe to be above the target vertebra.* + +| | | +| :--- | :--- | +| **Total Trajectories** | `[1024]` | +| **Total Hours** | `[]` | +| **Data Type** | `[ ] Clinical` `[] Ex-Vivo` `[ ] Table-Top Phantom` `[x] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +- Navigate ultrasound prove to find a specific anatomy + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [x] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[x] N/A` | +| **Collection Period** | From `[2025-03]` to `[2026-01]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + +**If yes, please briefly describe the recovery process:** + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [x] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +We start the ultrasound scan from various different initial positions. We scan multiple bone structures including vertebra L1-L5. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +*Kuka med14* + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `[None]` | +| **Room/3rd Person Camera** | `[None]` | +| **Force/Torque Sensor** | `[None]` | +| **Medical Imager** | `[e.g., Synthetic ultrasound by GAN, B-Mode]` | +| **Other** | `[Specify]` | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [x] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `None`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [x] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* +``` +action: [x, y, z, qw, qx, qy, qz] +- x, y, z: relative position in ultrasound image frame (meters) to next pose. +- qw, qx, qy, qz: quaternion rotation to the next frame +- Note that this 3D motion is along the surface of the training patient model. This may not align with the testing case. +``` + + +**Example:** +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +``` +observation.state.ee_state: [x, y, z, qw, qx, qy, qz] +- x, y, z: Absolute position in base frame (meters) +- qw, qx, qy, qz: quaternion in base frame +observation.state.joint_state: [j1, j2, j3, j4, j5, j6, j7] +- Absolute joint positions for 7-DOF arm (radians) +``` + + +**Example:** +``` +observation.state: [j1, j2, j3, j4, j5, j6, j7, gripper_pos] +- j1-j7: Absolute joint positions for 7-DOF arm (radians) +- gripper_pos: Current gripper opening (meters) +``` + +### πŸ“‹ Recommended Additional Representations + +*Even if not your primary action/state representation, we strongly encourage including these standardized formats for maximum compatibility:* + +**Recommended Action Fields:** +- **`action.cartesian_absolute`**: Absolute Cartesian pose with absolute quaternions + ``` + [x, y, z, qx, qy, qz, qw, gripper_angle] + ``` + +**Recommended State Fields:** +- **`observation.state.joint_positions`**: Absolute positions for all articulated joints + ``` + [joint_1, joint_2, ..., joint_n] + ``` + + +--- + +## ⏱️ Data Synchronization Approach + + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +*The synchronization is ensured automatically through simulation*. + +**Example:** *We collect joint kinematics from our Franka Research 3 and RGB-D frames from Intel RealSense D435 cameras, all running in ROS 2 Galactic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /camera/*/image_raw, and /camera/*/camera_info in a single rosbag2 session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±2 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Yunke Ao, Luohong Wu]` | +| **Institution** | `[Balgrist University Hospital]` | +| **Contact Email** | `[yunke.ao@balgrist.ch, luohong.wu@balgrist.ch, ...]` | +| **Citation (BibTeX)** | 
@misc{[Ultrabones100k_lerobot_2026],
  author = {[Yunke Ao, Luohong Wu, Philipp Fuernstahl]},
  title = {[Ultrabones100k Lerobot Dataset]},
  year = {2026},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Ultrasound/balgrist/sonogym_open_h_us_guidance_l4/README.md b/Ultrasound/balgrist/sonogym_open_h_us_guidance_l4/README.md new file mode 100644 index 0000000000000000000000000000000000000000..94fe78d030ac396140f6d88d92e192404b816dcf --- /dev/null +++ b/Ultrasound/balgrist/sonogym_open_h_us_guidance_l4/README.md @@ -0,0 +1,231 @@ + + +# SonoGym Probe Manipulation Lerobot Dataset_4 - README + +--- + +## πŸ“‹ At a Glance + +*Synthetic ultrasound probe manipulation to see L4 vertebra.* + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 1024 trajectories of expert policies to move the probe to be above the target vertebra.* + +| | | +| :--- | :--- | +| **Total Trajectories** | `[1024]` | +| **Total Hours** | `[]` | +| **Data Type** | `[ ] Clinical` `[] Ex-Vivo` `[ ] Table-Top Phantom` `[x] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +- Navigate ultrasound prove to find a specific anatomy + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [x] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[x] N/A` | +| **Collection Period** | From `[2025-03]` to `[2026-01]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + +**If yes, please briefly describe the recovery process:** + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [x] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +We start the ultrasound scan from various different initial positions. We scan multiple bone structures including vertebra L1-L5. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +*Kuka med14* + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `[None]` | +| **Room/3rd Person Camera** | `[None]` | +| **Force/Torque Sensor** | `[None]` | +| **Medical Imager** | `[e.g., Synthetic ultrasound by GAN, B-Mode]` | +| **Other** | `[Specify]` | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [x] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `None`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [x] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* +``` +action: [x, y, z, qw, qx, qy, qz] +- x, y, z: relative position in ultrasound image frame (meters) to next pose. +- qw, qx, qy, qz: quaternion rotation to the next frame +- Note that this 3D motion is along the surface of the training patient model. This may not align with the testing case. +``` + + +**Example:** +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +``` +observation.state.ee_state: [x, y, z, qw, qx, qy, qz] +- x, y, z: Absolute position in base frame (meters) +- qw, qx, qy, qz: quaternion in base frame +observation.state.joint_state: [j1, j2, j3, j4, j5, j6, j7] +- Absolute joint positions for 7-DOF arm (radians) +``` + + +**Example:** +``` +observation.state: [j1, j2, j3, j4, j5, j6, j7, gripper_pos] +- j1-j7: Absolute joint positions for 7-DOF arm (radians) +- gripper_pos: Current gripper opening (meters) +``` + +### πŸ“‹ Recommended Additional Representations + +*Even if not your primary action/state representation, we strongly encourage including these standardized formats for maximum compatibility:* + +**Recommended Action Fields:** +- **`action.cartesian_absolute`**: Absolute Cartesian pose with absolute quaternions + ``` + [x, y, z, qx, qy, qz, qw, gripper_angle] + ``` + +**Recommended State Fields:** +- **`observation.state.joint_positions`**: Absolute positions for all articulated joints + ``` + [joint_1, joint_2, ..., joint_n] + ``` + + +--- + +## ⏱️ Data Synchronization Approach + + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +*The synchronization is ensured automatically through simulation*. + +**Example:** *We collect joint kinematics from our Franka Research 3 and RGB-D frames from Intel RealSense D435 cameras, all running in ROS 2 Galactic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /camera/*/image_raw, and /camera/*/camera_info in a single rosbag2 session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±2 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Yunke Ao, Luohong Wu]` | +| **Institution** | `[Balgrist University Hospital]` | +| **Contact Email** | `[yunke.ao@balgrist.ch, luohong.wu@balgrist.ch, ...]` | +| **Citation (BibTeX)** | 
@misc{[Ultrabones100k_lerobot_2026],
  author = {[Yunke Ao, Luohong Wu, Philipp Fuernstahl]},
  title = {[Ultrabones100k Lerobot Dataset]},
  year = {2026},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Ultrasound/balgrist/sonogym_open_h_us_guidance_l5/README.md b/Ultrasound/balgrist/sonogym_open_h_us_guidance_l5/README.md new file mode 100644 index 0000000000000000000000000000000000000000..0781672cc3c9e6272f95aeb4bc235fbf2ce538ff --- /dev/null +++ b/Ultrasound/balgrist/sonogym_open_h_us_guidance_l5/README.md @@ -0,0 +1,231 @@ + + +# SonoGym Probe Manipulation Lerobot Dataset_5 - README + +--- + +## πŸ“‹ At a Glance + +*Synthetic ultrasound probe manipulation to see L5 vertebra.* + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 1024 trajectories of expert policies to move the probe to be above the target vertebra.* + +| | | +| :--- | :--- | +| **Total Trajectories** | `[1024]` | +| **Total Hours** | `[]` | +| **Data Type** | `[ ] Clinical` `[] Ex-Vivo` `[ ] Table-Top Phantom` `[x] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +- Navigate ultrasound prove to find a specific anatomy + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [x] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[x] N/A` | +| **Collection Period** | From `[2025-03]` to `[2026-01]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + +**If yes, please briefly describe the recovery process:** + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [x] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +We start the ultrasound scan from various different initial positions. We scan multiple bone structures including vertebra L1-L5. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +*Kuka med14* + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `[None]` | +| **Room/3rd Person Camera** | `[None]` | +| **Force/Torque Sensor** | `[None]` | +| **Medical Imager** | `[e.g., Synthetic ultrasound by GAN, B-Mode]` | +| **Other** | `[Specify]` | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [x] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `None`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [x] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* +``` +action: [x, y, z, qw, qx, qy, qz] +- x, y, z: relative position in ultrasound image frame (meters) to next pose. +- qw, qx, qy, qz: quaternion rotation to the next frame +- Note that this 3D motion is along the surface of the training patient model. This may not align with the testing case. +``` + + +**Example:** +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +``` +observation.state.ee_state: [x, y, z, qw, qx, qy, qz] +- x, y, z: Absolute position in base frame (meters) +- qw, qx, qy, qz: quaternion in base frame +observation.state.joint_state: [j1, j2, j3, j4, j5, j6, j7] +- Absolute joint positions for 7-DOF arm (radians) +``` + + +**Example:** +``` +observation.state: [j1, j2, j3, j4, j5, j6, j7, gripper_pos] +- j1-j7: Absolute joint positions for 7-DOF arm (radians) +- gripper_pos: Current gripper opening (meters) +``` + +### πŸ“‹ Recommended Additional Representations + +*Even if not your primary action/state representation, we strongly encourage including these standardized formats for maximum compatibility:* + +**Recommended Action Fields:** +- **`action.cartesian_absolute`**: Absolute Cartesian pose with absolute quaternions + ``` + [x, y, z, qx, qy, qz, qw, gripper_angle] + ``` + +**Recommended State Fields:** +- **`observation.state.joint_positions`**: Absolute positions for all articulated joints + ``` + [joint_1, joint_2, ..., joint_n] + ``` + + +--- + +## ⏱️ Data Synchronization Approach + + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +*The synchronization is ensured automatically through simulation*. + +**Example:** *We collect joint kinematics from our Franka Research 3 and RGB-D frames from Intel RealSense D435 cameras, all running in ROS 2 Galactic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /camera/*/image_raw, and /camera/*/camera_info in a single rosbag2 session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±2 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Yunke Ao, Luohong Wu]` | +| **Institution** | `[Balgrist University Hospital]` | +| **Contact Email** | `[yunke.ao@balgrist.ch, luohong.wu@balgrist.ch, ...]` | +| **Citation (BibTeX)** | 
@misc{[Ultrabones100k_lerobot_2026],
  author = {[Yunke Ao, Luohong Wu, Philipp Fuernstahl]},
  title = {[Ultrabones100k Lerobot Dataset]},
  year = {2026},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Ultrasound/balgrist/ultrabones_lerobot_dataset_full/README.md b/Ultrasound/balgrist/ultrabones_lerobot_dataset_full/README.md new file mode 100644 index 0000000000000000000000000000000000000000..388221dffe3bc469886bf183723f9473f9c0a8b2 --- /dev/null +++ b/Ultrasound/balgrist/ultrabones_lerobot_dataset_full/README.md @@ -0,0 +1,226 @@ + + +# Ultrabones100k Cadaveric Ultrasound Lerobot Dataset_1 - README + +--- + +## πŸ“‹ At a Glance + +*Freehand tracked ultrasound scan from an expert surgeon (first part ), with 14 cadavers and 5-12 different scans for each of them. Both Tibia and Fibula are scanned for different cadavers.* + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 60 trajectories of expert surgeons using the ultrasound probe to scan bones. The state is the current pose. Observation is the ultrasound image. The action is extracted as the delta cartesion pose to 1 second later.* + +| | | +| :--- | :--- | +| **Total Trajectories** | `[60]` | +| **Total Hours** | `[1.2]` | +| **Data Type** | `[ ] Clinical` `[x] Ex-Vivo` `[ ] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +- Ultrasound scanning for orthopedics / reconstruction (fibula and tibia) + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [x] **Other** (Please specify: `[Expert freehand]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[x] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2024-01]` to `[2024-06]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + +**If yes, please briefly describe the recovery process:** + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [x] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [x] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +We start the ultrasound scan from different initial positions. We scan multiple bone structures including fibla and tibia. We scan 14 different cadavers with 5-12 various recordings for each of them. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `[None]` | +| **Room/3rd Person Camera** | `[None]` | +| **Force/Torque Sensor** | `[None]` | +| **Medical Imager** | `[e.g., SuperSonic Imagine SL18-5 Ultrasound, B-Mode]` | +| **Other** | `[Specify]` | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [x] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [ ] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [x] **Other** (Please specify: `extrinsic xyz euler angle`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [x] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* +``` +action: [x, y, z, euler_x, euler_y, euler_z] +- x, y, z: relative position in ultrasound image frame (meters) to 1 sec later +- euler_x, euler_y, euler_z: extrinsic xyz euler angle in image frame to 1 sec later +``` + + +**Example:** +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +``` +observation.state: [x, y, z, euler_x, euler_y, euler_z] +- x, y, z: Absolute position in world frame (meters) +- euler_x, euler_y, euler_z: extrinsic xyz euler angle XYZ euler angle in world frame +``` + + +**Example:** +``` +observation.state: [j1, j2, j3, j4, j5, j6, j7, gripper_pos] +- j1-j7: Absolute joint positions for 7-DOF arm (radians) +- gripper_pos: Current gripper opening (meters) +``` + +### πŸ“‹ Recommended Additional Representations + +*Even if not your primary action/state representation, we strongly encourage including these standardized formats for maximum compatibility:* + +**Recommended Action Fields:** +- **`action.cartesian_absolute`**: Absolute Cartesian pose with absolute quaternions + ``` + [x, y, z, qx, qy, qz, qw, gripper_angle] + ``` + +**Recommended State Fields:** +- **`observation.state.joint_positions`**: Absolute positions for all articulated joints + ``` + [joint_1, joint_2, ..., joint_n] + ``` + + +--- + +## ⏱️ Data Synchronization Approach + + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +*The synchronization is part of of calibration procedure, which is detailed in [Ultrabones100k](https://arxiv.org/abs/2502.03783)*. + +**Example:** *We collect joint kinematics from our Franka Research 3 and RGB-D frames from Intel RealSense D435 cameras, all running in ROS 2 Galactic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /camera/*/image_raw, and /camera/*/camera_info in a single rosbag2 session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±2 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Yunke Ao, Luohong Wu]` | +| **Institution** | `[Balgrist University Hospital]` | +| **Contact Email** | `[yunke.ao@balgrist.ch, luohong.wu@balgrist.ch, ...]` | +| **Citation (BibTeX)** | 
@misc{[Ultrabones100k_lerobot_2026],
  author = {[Yunke Ao, Luohong Wu, Philipp Fuernstahl]},
  title = {[Ultrabones100k Lerobot Dataset]},
  year = {2026},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Ultrasound/balgrist/ultrabones_lerobot_dataset_full_2/README.md b/Ultrasound/balgrist/ultrabones_lerobot_dataset_full_2/README.md new file mode 100644 index 0000000000000000000000000000000000000000..eaf69fb5df76eafc5301f702c0b0c39bc22048cc --- /dev/null +++ b/Ultrasound/balgrist/ultrabones_lerobot_dataset_full_2/README.md @@ -0,0 +1,226 @@ + + +# Ultrabones100k Cadaveric Ultrasound Lerobot Dataset_2 - README + +--- + +## πŸ“‹ At a Glance + +*Freehand tracked ultrasound scan from an expert surgeon (second part), with 14 cadavers and 5-12 different scans for each of them. Both Tibia and Fibula are scanned for different cadavers.* + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 12 trajectories of expert surgeons using the ultrasound probe to scan bones, with different ultrasound image shape to the first one. The state is the current pose. Observation is the ultrasound image. The action is extracted as the delta cartesion pose to 1 second later.* + +| | | +| :--- | :--- | +| **Total Trajectories** | `[12]` | +| **Total Hours** | `[0.24]` | +| **Data Type** | `[ ] Clinical` `[x] Ex-Vivo` `[ ] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +- Ultrasound scanning for orthopedics / reconstruction (fibula and tibia) + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [x] **Other** (Please specify: `[Expert freehand]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[x] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2024-01]` to `[2024-06]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + +**If yes, please briefly describe the recovery process:** + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [x] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [x] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + + We start the ultrasound scan from different initial positions. We scan multiple bone structures including fibla and tibia. We scan 14 different cadavers with 5-12 various recordings for each of them. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `[None]` | +| **Room/3rd Person Camera** | `[None]` | +| **Force/Torque Sensor** | `[None]` | +| **Medical Imager** | `[e.g., SuperSonic Imagine SL18-5 Ultrasound, B-Mode]` | +| **Other** | `[Specify]` | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [x] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [ ] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [x] **Other** (Please specify: `extrinsic XYZ euler angle`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [x] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* +``` +action: [x, y, z, euler_x, euler_y, euler_z] +- x, y, z: relative position in ultrasound image frame (meters) to 1 sec later +- euler_x, euler_y, euler_z: extrinsic XYZ euler angle in image frame to 1 sec later +``` + + +**Example:** +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +``` +observation.state: [x, y, z, euler_x, euler_y, euler_z] +- x, y, z: Absolute position in world frame (meters) +- euler_x, euler_y, euler_z: extrinsic XYZ euler angle in world frame +``` + + +**Example:** +``` +observation.state: [j1, j2, j3, j4, j5, j6, j7, gripper_pos] +- j1-j7: Absolute joint positions for 7-DOF arm (radians) +- gripper_pos: Current gripper opening (meters) +``` + +### πŸ“‹ Recommended Additional Representations + +*Even if not your primary action/state representation, we strongly encourage including these standardized formats for maximum compatibility:* + +**Recommended Action Fields:** +- **`action.cartesian_absolute`**: Absolute Cartesian pose with absolute quaternions + ``` + [x, y, z, qx, qy, qz, qw, gripper_angle] + ``` + +**Recommended State Fields:** +- **`observation.state.joint_positions`**: Absolute positions for all articulated joints + ``` + [joint_1, joint_2, ..., joint_n] + ``` + + +--- + +## ⏱️ Data Synchronization Approach + + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +*The synchronization is part of of calibration procedure, which is detailed in [Ultrabones100k](https://arxiv.org/abs/2502.03783)*. + +**Example:** *We collect joint kinematics from our Franka Research 3 and RGB-D frames from Intel RealSense D435 cameras, all running in ROS 2 Galactic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /camera/*/image_raw, and /camera/*/camera_info in a single rosbag2 session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±2 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Yunke Ao, Luohong Wu]` | +| **Institution** | `[Balgrist University Hospital]` | +| **Contact Email** | `[yunke.ao@balgrist.ch, luohong.wu@balgrist.ch, ...]` | +| **Citation (BibTeX)** | 
@misc{[Ultrabones100k_lerobot_2026],
  author = {[Yunke Ao, Luohong Wu, Philipp Fuernstahl]},
  title = {[Ultrabones100k Lerobot Dataset]},
  year = {2026},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Ultrasound/balgrist/ultrabones_lerobot_dataset_full_2_synthetic_robot_2/README.md b/Ultrasound/balgrist/ultrabones_lerobot_dataset_full_2_synthetic_robot_2/README.md new file mode 100644 index 0000000000000000000000000000000000000000..48466b7433d4db4d6abe033a81713142b34a2bee --- /dev/null +++ b/Ultrasound/balgrist/ultrabones_lerobot_dataset_full_2_synthetic_robot_2/README.md @@ -0,0 +1,228 @@ + + +# Ultrabones100k Cadaveric Ultrasound Dataset with synthetic robot state_2- README + +--- + +## πŸ“‹ At a Glance + +*Freehand tracked ultrasound scan from an expert surgeon, augmented with synthetic robot state with randomized base positions and ee to us extrinsics.* + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 12 trajectories of expert surgeons using the ultrasound probe to scan bones, with different ultrasound image shape to the second one. The state is the current pose. Observation is the ultrasound image. The action is extracted as the delta cartesion pose to 1 second later.* + +| | | +| :--- | :--- | +| **Total Trajectories** | `[12]` | +| **Total Hours** | `[0.24]` | +| **Data Type** | `[ ] Clinical` `[x] Ex-Vivo` `[ ] Table-Top Phantom` `[x] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +- Ultrasound scanning for orthopedics / reconstruction (fibula and tibia) + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [x] **Other** (Please specify: `[Expert freehand]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[x] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2024-01]` to `[2024-06]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + +**If yes, please briefly describe the recovery process:** + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [x] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [x] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + + We start the ultrasound scan from different initial positions. We scan multiple bone structures including fibla and tibia. We scan 14 different cadavers with 5-12 various recordings for each of them. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*Kuka med14* + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `[None]` | +| **Room/3rd Person Camera** | `[None]` | +| **Force/Torque Sensor** | `[None]` | +| **Medical Imager** | `[e.g., SuperSonic Imagine SL18-5 Ultrasound, B-Mode]` | +| **Other** | `[Specify]` | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [x] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `None`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [x] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* +``` +action: [x, y, z, qw, qx, qy, qz] +- x, y, z: relative position in ultrasound image frame (meters) to 1 sec later +- qw, qx, qy, qz: quaternion in image frame to 1 sec later +``` + + +**Example:** +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +``` +observation.state.us_pose: [x, y, z, qw, qx, qy, qz] +- x, y, z: Absolute position in robot base frame (meters) +- qw, qx, qy, qz: quaternion in robot base frame +observation.state.joint_pos: [j1, j2, ..., j7] +- j1-j7: joint positions +``` + + +**Example:** +``` +observation.state: [j1, j2, j3, j4, j5, j6, j7, gripper_pos] +- j1-j7: Absolute joint positions for 7-DOF arm (radians) +- gripper_pos: Current gripper opening (meters) +``` + +### πŸ“‹ Recommended Additional Representations + +*Even if not your primary action/state representation, we strongly encourage including these standardized formats for maximum compatibility:* + +**Recommended Action Fields:** +- **`action.cartesian_absolute`**: Absolute Cartesian pose with absolute quaternions + ``` + [x, y, z, qx, qy, qz, qw, gripper_angle] + ``` + +**Recommended State Fields:** +- **`observation.state.joint_positions`**: Absolute positions for all articulated joints + ``` + [joint_1, joint_2, ..., joint_n] + ``` + + +--- + +## ⏱️ Data Synchronization Approach + + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +*The synchronization is part of of calibration procedure, which is detailed in [Ultrabones100k](https://arxiv.org/abs/2502.03783)*. + +**Example:** *We collect joint kinematics from our Franka Research 3 and RGB-D frames from Intel RealSense D435 cameras, all running in ROS 2 Galactic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /camera/*/image_raw, and /camera/*/camera_info in a single rosbag2 session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±2 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Yunke Ao, Luohong Wu]` | +| **Institution** | `[Balgrist University Hospital]` | +| **Contact Email** | `[yunke.ao@balgrist.ch, luohong.wu@balgrist.ch, ...]` | +| **Citation (BibTeX)** | 
@misc{[Ultrabones100k_lerobot_2026],
  author = {[Yunke Ao, Luohong Wu, Philipp Fuernstahl]},
  title = {[Ultrabones100k Lerobot Dataset]},
  year = {2026},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Ultrasound/balgrist/ultrabones_lerobot_dataset_full_3/README.md b/Ultrasound/balgrist/ultrabones_lerobot_dataset_full_3/README.md new file mode 100644 index 0000000000000000000000000000000000000000..489e3d72dc89309845a62b57b7ba29767f2da8af --- /dev/null +++ b/Ultrasound/balgrist/ultrabones_lerobot_dataset_full_3/README.md @@ -0,0 +1,226 @@ + + +# Ultrabones100k Cadaveric Ultrasound Lerobot Dataset_3 - README + +--- + +## πŸ“‹ At a Glance + +*Freehand tracked ultrasound scan from an expert surgeon (third part), with 14 cadavers and 5-12 different scans for each of them. Both Tibia and Fibula are scanned for different cadavers.* + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 18 trajectories of expert surgeons using the ultrasound probe to scan bones, with different ultrasound image shape to the second one. The state is the current pose. Observation is the ultrasound image. The action is extracted as the delta cartesion pose to 1 second later.* + +| | | +| :--- | :--- | +| **Total Trajectories** | `[18]` | +| **Total Hours** | `[0.36]` | +| **Data Type** | `[ ] Clinical` `[x] Ex-Vivo` `[ ] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +- Ultrasound scanning for orthopedics / reconstruction (fibula and tibia) + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [x] **Other** (Please specify: `[Expert freehand]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[x] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2024-01]` to `[2024-06]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + +**If yes, please briefly describe the recovery process:** + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [x] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [x] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + + We start the ultrasound scan from different initial positions. We scan multiple bone structures including fibla and tibia. We scan 14 different cadavers with 5-12 various recordings for each of them. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `[None]` | +| **Room/3rd Person Camera** | `[None]` | +| **Force/Torque Sensor** | `[None]` | +| **Medical Imager** | `[e.g., SuperSonic Imagine SL18-5 Ultrasound, B-Mode]` | +| **Other** | `[Specify]` | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [x] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [ ] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [x] **Other** (Please specify: `extrinsic XYZ euler angle`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [x] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* +``` +action: [x, y, z, euler_x, euler_y, euler_z] +- x, y, z: relative position in ultrasound image frame (meters) to 1 sec later +- euler_x, euler_y, euler_z: extrinsic XYZ euler angle in image frame to 1 sec later +``` + + +**Example:** +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +``` +observation.state: [x, y, z, euler_x, euler_y, euler_z] +- x, y, z: Absolute position in world frame (meters) +- euler_x, euler_y, euler_z: extrinsic XYZ euler angle in world frame +``` + + +**Example:** +``` +observation.state: [j1, j2, j3, j4, j5, j6, j7, gripper_pos] +- j1-j7: Absolute joint positions for 7-DOF arm (radians) +- gripper_pos: Current gripper opening (meters) +``` + +### πŸ“‹ Recommended Additional Representations + +*Even if not your primary action/state representation, we strongly encourage including these standardized formats for maximum compatibility:* + +**Recommended Action Fields:** +- **`action.cartesian_absolute`**: Absolute Cartesian pose with absolute quaternions + ``` + [x, y, z, qx, qy, qz, qw, gripper_angle] + ``` + +**Recommended State Fields:** +- **`observation.state.joint_positions`**: Absolute positions for all articulated joints + ``` + [joint_1, joint_2, ..., joint_n] + ``` + + +--- + +## ⏱️ Data Synchronization Approach + + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +*The synchronization is part of of calibration procedure, which is detailed in [Ultrabones100k](https://arxiv.org/abs/2502.03783)*. + +**Example:** *We collect joint kinematics from our Franka Research 3 and RGB-D frames from Intel RealSense D435 cameras, all running in ROS 2 Galactic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /camera/*/image_raw, and /camera/*/camera_info in a single rosbag2 session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±2 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Yunke Ao, Luohong Wu]` | +| **Institution** | `[Balgrist University Hospital]` | +| **Contact Email** | `[yunke.ao@balgrist.ch, luohong.wu@balgrist.ch, ...]` | +| **Citation (BibTeX)** | 
@misc{[Ultrabones100k_lerobot_2026],
  author = {[Yunke Ao, Luohong Wu, Philipp Fuernstahl]},
  title = {[Ultrabones100k Lerobot Dataset]},
  year = {2026},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Ultrasound/balgrist/ultrabones_lerobot_dataset_full_3_synthetic_robot_2/README.md b/Ultrasound/balgrist/ultrabones_lerobot_dataset_full_3_synthetic_robot_2/README.md new file mode 100644 index 0000000000000000000000000000000000000000..15156627205f91bda4a87a550f7eefb0d6a0354e --- /dev/null +++ b/Ultrasound/balgrist/ultrabones_lerobot_dataset_full_3_synthetic_robot_2/README.md @@ -0,0 +1,228 @@ + + +# Ultrabones100k Cadaveric Ultrasound Dataset with synthetic robot state_3- README + +--- + +## πŸ“‹ At a Glance + +*Freehand tracked ultrasound scan from an expert surgeon, augmented with synthetic robot state with randomized base positions and ee to us extrinsics.* + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 18 trajectories of expert surgeons using the ultrasound probe to scan bones, with different ultrasound image shape to the second one. The state is the current pose. Observation is the ultrasound image. The action is extracted as the delta cartesion pose to 1 second later.* + +| | | +| :--- | :--- | +| **Total Trajectories** | `[18]` | +| **Total Hours** | `[0.36]` | +| **Data Type** | `[ ] Clinical` `[x] Ex-Vivo` `[ ] Table-Top Phantom` `[x] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +- Ultrasound scanning for orthopedics / reconstruction (fibula and tibia) + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [x] **Other** (Please specify: `[Expert freehand]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[x] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2024-01]` to `[2024-06]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + +**If yes, please briefly describe the recovery process:** + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [x] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [x] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + + We start the ultrasound scan from different initial positions. We scan multiple bone structures including fibla and tibia. We scan 14 different cadavers with 5-12 various recordings for each of them. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*Kuka med14* + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `[None]` | +| **Room/3rd Person Camera** | `[None]` | +| **Force/Torque Sensor** | `[None]` | +| **Medical Imager** | `[e.g., SuperSonic Imagine SL18-5 Ultrasound, B-Mode]` | +| **Other** | `[Specify]` | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [x] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `None`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [x] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* +``` +action: [x, y, z, qw, qx, qy, qz] +- x, y, z: relative position in ultrasound image frame (meters) to 1 sec later +- qw, qx, qy, qz: quaternion in image frame to 1 sec later +``` + + +**Example:** +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +``` +observation.state.us_pose: [x, y, z, qw, qx, qy, qz] +- x, y, z: Absolute position in robot base frame (meters) +- qw, qx, qy, qz: quaternion in robot base frame +observation.state.joint_pos: [j1, j2, ..., j7] +- j1-j7: joint positions +``` + + +**Example:** +``` +observation.state: [j1, j2, j3, j4, j5, j6, j7, gripper_pos] +- j1-j7: Absolute joint positions for 7-DOF arm (radians) +- gripper_pos: Current gripper opening (meters) +``` + +### πŸ“‹ Recommended Additional Representations + +*Even if not your primary action/state representation, we strongly encourage including these standardized formats for maximum compatibility:* + +**Recommended Action Fields:** +- **`action.cartesian_absolute`**: Absolute Cartesian pose with absolute quaternions + ``` + [x, y, z, qx, qy, qz, qw, gripper_angle] + ``` + +**Recommended State Fields:** +- **`observation.state.joint_positions`**: Absolute positions for all articulated joints + ``` + [joint_1, joint_2, ..., joint_n] + ``` + + +--- + +## ⏱️ Data Synchronization Approach + + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +*The synchronization is part of of calibration procedure, which is detailed in [Ultrabones100k](https://arxiv.org/abs/2502.03783)*. + +**Example:** *We collect joint kinematics from our Franka Research 3 and RGB-D frames from Intel RealSense D435 cameras, all running in ROS 2 Galactic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /camera/*/image_raw, and /camera/*/camera_info in a single rosbag2 session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±2 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Yunke Ao, Luohong Wu]` | +| **Institution** | `[Balgrist University Hospital]` | +| **Contact Email** | `[yunke.ao@balgrist.ch, luohong.wu@balgrist.ch, ...]` | +| **Citation (BibTeX)** | 
@misc{[Ultrabones100k_lerobot_2026],
  author = {[Yunke Ao, Luohong Wu, Philipp Fuernstahl]},
  title = {[Ultrabones100k Lerobot Dataset]},
  year = {2026},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Ultrasound/balgrist/ultrabones_lerobot_dataset_full_synthetic_robot/README.md b/Ultrasound/balgrist/ultrabones_lerobot_dataset_full_synthetic_robot/README.md new file mode 100644 index 0000000000000000000000000000000000000000..c8eacee68dbeba95a8a8a57eaff55cb140cdde5e --- /dev/null +++ b/Ultrasound/balgrist/ultrabones_lerobot_dataset_full_synthetic_robot/README.md @@ -0,0 +1,228 @@ + + +# Ultrabones100k Cadaveric Ultrasound Dataset with synthetic robot state- README + +--- + +## πŸ“‹ At a Glance + +*Freehand tracked ultrasound scan from an expert surgeon, augmented with synthetic robot state with randomized base positions and ee to us extrinsics.* + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 60 trajectories of expert surgeons using the ultrasound probe to scan bones, with different ultrasound image shape to the second one. The state is the current pose. Observation is the ultrasound image. The action is extracted as the delta cartesion pose to 1 second later.* + +| | | +| :--- | :--- | +| **Total Trajectories** | `[60]` | +| **Total Hours** | `[1.2]` | +| **Data Type** | `[ ] Clinical` `[x] Ex-Vivo` `[ ] Table-Top Phantom` `[x] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +- Ultrasound scanning for orthopedics / reconstruction (fibula and tibia) + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [x] **Other** (Please specify: `[Expert freehand]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[x] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2024-01]` to `[2024-06]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + +**If yes, please briefly describe the recovery process:** + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [x] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [x] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + + We start the ultrasound scan from different initial positions. We scan multiple bone structures including fibla and tibia. We scan 14 different cadavers with 5-12 various recordings for each of them. + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*Kuka med14* + + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Primary Camera** | `[None]` | +| **Room/3rd Person Camera** | `[None]` | +| **Force/Torque Sensor** | `[None]` | +| **Medical Imager** | `[e.g., SuperSonic Imagine SL18-5 Ultrasound, B-Mode]` | +| **Other** | `[Specify]` | + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [ ] **Absolute Cartesian** (position/orientation relative to robot base) +- [x] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `None`) + +**Reference Frame:** +- [ ] **Robot Base Frame** +- [x] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + +**Action Dimensions:** +*List the action space dimensions and their meanings.* +``` +action: [x, y, z, qw, qx, qy, qz] +- x, y, z: relative position in ultrasound image frame (meters) to 1 sec later +- qw, qx, qy, qz: quaternion in image frame to 1 sec later +``` + + +**Example:** +``` +action: [x, y, z, qx, qy, qz, qw, gripper] +- x, y, z: Absolute position in robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +- gripper: Gripper opening angle (radians) +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + +**State Dimensions:** +*List the state space dimensions and their meanings.* + +``` +observation.state.us_pose: [x, y, z, qw, qx, qy, qz] +- x, y, z: Absolute position in robot base frame (meters) +- qw, qx, qy, qz: quaternion in robot base frame +observation.state.joint_pos: [j1, j2, ..., j7] +- j1-j7: joint positions +``` + + +**Example:** +``` +observation.state: [j1, j2, j3, j4, j5, j6, j7, gripper_pos] +- j1-j7: Absolute joint positions for 7-DOF arm (radians) +- gripper_pos: Current gripper opening (meters) +``` + +### πŸ“‹ Recommended Additional Representations + +*Even if not your primary action/state representation, we strongly encourage including these standardized formats for maximum compatibility:* + +**Recommended Action Fields:** +- **`action.cartesian_absolute`**: Absolute Cartesian pose with absolute quaternions + ``` + [x, y, z, qx, qy, qz, qw, gripper_angle] + ``` + +**Recommended State Fields:** +- **`observation.state.joint_positions`**: Absolute positions for all articulated joints + ``` + [joint_1, joint_2, ..., joint_n] + ``` + + +--- + +## ⏱️ Data Synchronization Approach + + +*Describe how you achieved proper data synchronization across different sensors, cameras, and robotic systems during data collection. This is crucial for ensuring temporal alignment of all modalities in your dataset.* + +*The synchronization is part of of calibration procedure, which is detailed in [Ultrabones100k](https://arxiv.org/abs/2502.03783)*. + +**Example:** *We collect joint kinematics from our Franka Research 3 and RGB-D frames from Intel RealSense D435 cameras, all running in ROS 2 Galactic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /camera/*/image_raw, and /camera/*/camera_info in a single rosbag2 session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±2 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Yunke Ao, Luohong Wu]` | +| **Institution** | `[Balgrist University Hospital]` | +| **Contact Email** | `[yunke.ao@balgrist.ch, luohong.wu@balgrist.ch, ...]` | +| **Citation (BibTeX)** | 
@misc{[Ultrabones100k_lerobot_2026],
  author = {[Yunke Ao, Luohong Wu, Philipp Fuernstahl]},
  title = {[Ultrabones100k Lerobot Dataset]},
  year = {2026},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Ultrasound/cuhk/bernard_cheng/us_ppnr/needle_retrieval/README.md b/Ultrasound/cuhk/bernard_cheng/us_ppnr/needle_retrieval/README.md new file mode 100644 index 0000000000000000000000000000000000000000..742ac8f66f517bffd65d726349ad8d4b32033d3e --- /dev/null +++ b/Ultrasound/cuhk/bernard_cheng/us_ppnr/needle_retrieval/README.md @@ -0,0 +1,188 @@ +# A Multimodal Dataset for Autonomous Probe Placement and Needle Retrieval in Ultrasound-Guided Liver Biopsy (US-PPNR) - README + +--- + +## πŸ“‹ At a Glance + +Multimodal, time-synchronized demonstrations of ultrasound-guided needle biopsy on clinically realistic phantoms/tissue, including ultrasound imaging, multi-view video, robot kinematics, force/torque, and optical tracking. + +--- + +## πŸ“– Dataset Overview + +This dataset supports training and evaluating vision-language-action (VLA) and imitation-learning policies for **ultrasound (US)-guided needle biopsy**. It covers the procedures including: scanning and probe placement, needle tracking, and probe adjustments to maintain needle visualization. Each trajectory provides synchronized **US B-mode**, **RGB room video**, **RGB-D wrist video**, **robot joint/task-space states**, **force/torque**, and **NDI optical tracking** for needle/probe/camera poses, plus time-aligned narration/metadata. + +| | | +|:-----------------------|:-------------------------------------------------------------------------------------------------------| +| **Total Trajectories** | 2,546 | +| **Total Hours** | 10:51:09 | +| **Data Type** | \[βœ…] Clinical \[βœ…] Ex-Vivo \[βœ…] Table-Top Phantom \[ ] Digital Simulation \[βœ…] Physical Simulation | +| **License** | CC BY 4.0 | +| **Version** | 1.0 | + +--- + +## 🎯 Tasks & Domain + +### Domain + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- Probe Placement (**PP.tar.gz**): + - Autonomous probe placement for liver scanning +- Needle Retrieval (**NR.tar.gz**): + - Robust needle tracking during insertion + - Autonomous needle retrieval from phantom/tissue under US guidance + - Autonomous probe scanning (sweeps, plane finding) + - Adaptive probe manipulation to maintain visualization + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +- [x] **Human Teleoperation** +- [x] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** + + +### Operator Details + +| | Description | +|:-------------------------|:-------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Operator Count** | 6 | +| **Operator Skill Level** | \[x] Expert (e.g., Sonographer/Clinician)
\[x] Intermediate (e.g., Trained Researcher)
\[x] Novice (e.g., ML Researcher with minimal experience)
\[ ] N/A | +| **Collection Period** | From 2025-12-20 to 2026-01-18 | + +### Recovery Demonstrations + +- [ ] **Yes** +- [x] **No** + +--- + +## πŸ’‘ Diversity Dimensions + +- [x] **Camera Position / Angle** +- [x] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, tissue types, lesion locations) +- [x] **Spatial Layout** (e.g., varying lesion position and insertion approach) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [x] **Task Execution** (e.g., different scanning/insertion techniques) +- [x] **Background / Scene** +- [ ] **Other** + +**Elaboration:** + +Data were collected across both commercial and custom-built phantoms to vary tissue appearance, acoustic properties, and lesion presentation. +Trials span a range of target locations, scanning planes, insertion angles, and operator strategies (expert vs sub-expert), yielding diversity in both perception (US appearance, needle visibility) and control (probe/needle trajectories, contact forces). + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +- UR5e robotic arm for **probe manipulation** (end-effector-mounted force/torque sensing; tracked probe marker) + + +### Sensors & Cameras + +| Type | Model/Details | +|:---------------------------|:----------------------------------------------------------------------------------------------------------| +| **Medical Imager** | Ultrasound B-mode stream β€” **Wisonic Clover 60** + **Wisonic C5-1 convex transducer**, 1920Γ—1080 @ 30 FPS | +| **Room/3rd Person Camera** | **NDI Polaris Vega VT RGB camera**, 1024Γ—768 @ 30 FPS | +| **RGB-D Wrist Camera** | **ZED 2**, RGB + Depth, 640Γ—480 @ 30 FPS (mounted on probe end-effector) | +| **Force/Torque Sensor** | **ATI Axia80-M8**, 30 HZ, End-effector F/T sensor | +| **Optical Tracker** | **NDI Polaris Vega VT**, 30 Hz, tracked poses for needle/probe/cameras/components | + +--- + +## 🎯 Action & State Space Representation + +This dataset follows the **LeRobot** format and provides synchronized actions, robot state, imaging, and tracking. + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) +- [x] **Relative Cartesian** (delta position/orientation from current pose) *(typical for teleop step commands)* +- [x] **Joint Space** (direct joint commands recorded from robot interface) +- [ ] **Other** + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** + +**Reference Frame:** +- [x] **Robot Base Frame** +- [x] **Tool/End-Effector Frame** +- [x] **World/Global Frame** *(via optical tracker frame)* +- [ ] **Camera Frame** +- [x] **Other** (calibrated US image frame) + +**Action Dimensions:** + +Actions are provided per-controlled subsystem (probe manipulation and needle insertion). + +**Example:** +```text +action.probe_delta: "probe_delta_x", "probe_delta_y", "probe_delta_z", "probe_delta_ux", "probe_delta_uy", "probe_delta_uz", "probe_delta_w", +- "probe_delta_x", "probe_delta_y", "probe_delta_z": position in UR5e robot base frame (meters) +- "probe_delta_ux", "probe_delta_uy", "probe_delta_uz", "probe_delta_w": orientation as quaternion + +action.needle_delta: "needle_delta_x", "needle_delta_y", "needle_delta_z", "needle_delta_ux", "needle_delta_uy", "needle_delta_uz", "needle_delta_w" +- "needle_delta_x", "needle_delta_y", "needle_delta_z": position in NDI optical tracker frame (meters) +- "needle_delta_ux", "needle_delta_uy", "needle_delta_uz", "needle_delta_w": orientation as quaternion +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [x] **Force/Torque Readings** +- [ ] **Gripper State** +- [x] **Other** (optical tracker poses for needle/probe; calibrated transforms) + +**State Dimensions:** +```text +observation.state.joint_positions: "elbow_joint", "shoulder_lift_joint", "shoulder_pan_joint", "wrist_1_joint", "wrist_2_joint", "wrist_3_joint", +observation.state.ee_pose: "probe_ur_x", "probe_ur_y", "probe_ur_z", "probe_ur_ux", "probe_ur_uy", "probe_ur_uz", "probe_ur_w", +observation.state.ee_pose_ndi: "probe_ndi_x", "probe_ndi_y", "probe_ndi_z", "probe_ndi_ux", "probe_ndi_uy", "probe_ndi_uz", "probe_ndi_w", +observation.state.needle_pose_ndi: "needle_ndi_x", "needle_ndi_y", "needle_ndi_z", "needle_ndi_ux", "needle_ndi_uy", "needle_ndi_uz", "needle_ndi_w", +observation.meta.force_torque: "fx", "fy", "fz", "tx", "ty", "tz" +``` + +--- + +## ⏱️ Data Synchronization Approach + +All modalities are synchronized to a common time base and exported with corrected timestamps. + +- **Streams & rates:** robot kinematics and force/torque at **500 Hz**; NDI optical tracking at **60 Hz**; RGB/RGB-D/US at **30 FPS**. +- **Delay measurement:** the robot executes controlled sinusoidal motions while each sensor stream observes the motion (needle/probe/camera movement). Phase differences between each sensor’s measured sinusoid and the robot reference are used to estimate per-stream delays. The delay was calculated using [PLUS Toolkit](https://pmc.ncbi.nlm.nih.gov/articles/PMC4437531/). +- **Alignment:** timestamps are corrected using measured delays, and asynchronous streams are aligned via interpolation/resampling to produce unified, time-aligned trajectories. +- **Recording:** all raw streams and teleop inputs are recorded in ROS (rosbag). + +--- + +## πŸ‘₯ Attribution & Contact + +| | | +|:---------------------------|:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Dataset Lead** | Shing Shin Cheng; Wei Wang; Qingpeng Ding; Yuelin Zhang; Zhouyang Hong; Luoyao Kang; Wenxuan Xie | +| **Institution** | The Chinese University of Hong Kong (CUHK); First Affiliated Hospital, Sun Yat-Sen University | +| **Contact Email** | sscheng@cuhk.edu.hk | +| **Citation (BibTeX)** | 
@misc{openh_ausnb_2026,
  author = {Cheng, Shing Shin and Wang, Wei and Ding, Qingpeng and Zhang, Yuelin and Hong, Zhouyang and Kang, Luoyao and Xie, Wenxuan},
  title = {A Multimodal Dataset for Autonomous Probe Placement and Needle Retrieval in Ultrasound-Guided Liver Biopsy (US-PPNR)},
  year = {2026},
  publisher = {Open-H-Embodiment},
  license = {CC BY 4.0}, 
}
| +--- \ No newline at end of file diff --git a/Ultrasound/cuhk/bernard_cheng/us_ppnr/probe_placement/README.md b/Ultrasound/cuhk/bernard_cheng/us_ppnr/probe_placement/README.md new file mode 100644 index 0000000000000000000000000000000000000000..742ac8f66f517bffd65d726349ad8d4b32033d3e --- /dev/null +++ b/Ultrasound/cuhk/bernard_cheng/us_ppnr/probe_placement/README.md @@ -0,0 +1,188 @@ +# A Multimodal Dataset for Autonomous Probe Placement and Needle Retrieval in Ultrasound-Guided Liver Biopsy (US-PPNR) - README + +--- + +## πŸ“‹ At a Glance + +Multimodal, time-synchronized demonstrations of ultrasound-guided needle biopsy on clinically realistic phantoms/tissue, including ultrasound imaging, multi-view video, robot kinematics, force/torque, and optical tracking. + +--- + +## πŸ“– Dataset Overview + +This dataset supports training and evaluating vision-language-action (VLA) and imitation-learning policies for **ultrasound (US)-guided needle biopsy**. It covers the procedures including: scanning and probe placement, needle tracking, and probe adjustments to maintain needle visualization. Each trajectory provides synchronized **US B-mode**, **RGB room video**, **RGB-D wrist video**, **robot joint/task-space states**, **force/torque**, and **NDI optical tracking** for needle/probe/camera poses, plus time-aligned narration/metadata. + +| | | +|:-----------------------|:-------------------------------------------------------------------------------------------------------| +| **Total Trajectories** | 2,546 | +| **Total Hours** | 10:51:09 | +| **Data Type** | \[βœ…] Clinical \[βœ…] Ex-Vivo \[βœ…] Table-Top Phantom \[ ] Digital Simulation \[βœ…] Physical Simulation | +| **License** | CC BY 4.0 | +| **Version** | 1.0 | + +--- + +## 🎯 Tasks & Domain + +### Domain + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** + +### Demonstrated Skills + +- Probe Placement (**PP.tar.gz**): + - Autonomous probe placement for liver scanning +- Needle Retrieval (**NR.tar.gz**): + - Robust needle tracking during insertion + - Autonomous needle retrieval from phantom/tissue under US guidance + - Autonomous probe scanning (sweeps, plane finding) + - Adaptive probe manipulation to maintain visualization + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +- [x] **Human Teleoperation** +- [x] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** + + +### Operator Details + +| | Description | +|:-------------------------|:-------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Operator Count** | 6 | +| **Operator Skill Level** | \[x] Expert (e.g., Sonographer/Clinician)
\[x] Intermediate (e.g., Trained Researcher)
\[x] Novice (e.g., ML Researcher with minimal experience)
\[ ] N/A | +| **Collection Period** | From 2025-12-20 to 2026-01-18 | + +### Recovery Demonstrations + +- [ ] **Yes** +- [x] **No** + +--- + +## πŸ’‘ Diversity Dimensions + +- [x] **Camera Position / Angle** +- [x] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, tissue types, lesion locations) +- [x] **Spatial Layout** (e.g., varying lesion position and insertion approach) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [x] **Task Execution** (e.g., different scanning/insertion techniques) +- [x] **Background / Scene** +- [ ] **Other** + +**Elaboration:** + +Data were collected across both commercial and custom-built phantoms to vary tissue appearance, acoustic properties, and lesion presentation. +Trials span a range of target locations, scanning planes, insertion angles, and operator strategies (expert vs sub-expert), yielding diversity in both perception (US appearance, needle visibility) and control (probe/needle trajectories, contact forces). + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +- UR5e robotic arm for **probe manipulation** (end-effector-mounted force/torque sensing; tracked probe marker) + + +### Sensors & Cameras + +| Type | Model/Details | +|:---------------------------|:----------------------------------------------------------------------------------------------------------| +| **Medical Imager** | Ultrasound B-mode stream β€” **Wisonic Clover 60** + **Wisonic C5-1 convex transducer**, 1920Γ—1080 @ 30 FPS | +| **Room/3rd Person Camera** | **NDI Polaris Vega VT RGB camera**, 1024Γ—768 @ 30 FPS | +| **RGB-D Wrist Camera** | **ZED 2**, RGB + Depth, 640Γ—480 @ 30 FPS (mounted on probe end-effector) | +| **Force/Torque Sensor** | **ATI Axia80-M8**, 30 HZ, End-effector F/T sensor | +| **Optical Tracker** | **NDI Polaris Vega VT**, 30 Hz, tracked poses for needle/probe/cameras/components | + +--- + +## 🎯 Action & State Space Representation + +This dataset follows the **LeRobot** format and provides synchronized actions, robot state, imaging, and tracking. + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) +- [x] **Relative Cartesian** (delta position/orientation from current pose) *(typical for teleop step commands)* +- [x] **Joint Space** (direct joint commands recorded from robot interface) +- [ ] **Other** + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** + +**Reference Frame:** +- [x] **Robot Base Frame** +- [x] **Tool/End-Effector Frame** +- [x] **World/Global Frame** *(via optical tracker frame)* +- [ ] **Camera Frame** +- [x] **Other** (calibrated US image frame) + +**Action Dimensions:** + +Actions are provided per-controlled subsystem (probe manipulation and needle insertion). + +**Example:** +```text +action.probe_delta: "probe_delta_x", "probe_delta_y", "probe_delta_z", "probe_delta_ux", "probe_delta_uy", "probe_delta_uz", "probe_delta_w", +- "probe_delta_x", "probe_delta_y", "probe_delta_z": position in UR5e robot base frame (meters) +- "probe_delta_ux", "probe_delta_uy", "probe_delta_uz", "probe_delta_w": orientation as quaternion + +action.needle_delta: "needle_delta_x", "needle_delta_y", "needle_delta_z", "needle_delta_ux", "needle_delta_uy", "needle_delta_uz", "needle_delta_w" +- "needle_delta_x", "needle_delta_y", "needle_delta_z": position in NDI optical tracker frame (meters) +- "needle_delta_ux", "needle_delta_uy", "needle_delta_uz", "needle_delta_w": orientation as quaternion +``` + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [x] **Force/Torque Readings** +- [ ] **Gripper State** +- [x] **Other** (optical tracker poses for needle/probe; calibrated transforms) + +**State Dimensions:** +```text +observation.state.joint_positions: "elbow_joint", "shoulder_lift_joint", "shoulder_pan_joint", "wrist_1_joint", "wrist_2_joint", "wrist_3_joint", +observation.state.ee_pose: "probe_ur_x", "probe_ur_y", "probe_ur_z", "probe_ur_ux", "probe_ur_uy", "probe_ur_uz", "probe_ur_w", +observation.state.ee_pose_ndi: "probe_ndi_x", "probe_ndi_y", "probe_ndi_z", "probe_ndi_ux", "probe_ndi_uy", "probe_ndi_uz", "probe_ndi_w", +observation.state.needle_pose_ndi: "needle_ndi_x", "needle_ndi_y", "needle_ndi_z", "needle_ndi_ux", "needle_ndi_uy", "needle_ndi_uz", "needle_ndi_w", +observation.meta.force_torque: "fx", "fy", "fz", "tx", "ty", "tz" +``` + +--- + +## ⏱️ Data Synchronization Approach + +All modalities are synchronized to a common time base and exported with corrected timestamps. + +- **Streams & rates:** robot kinematics and force/torque at **500 Hz**; NDI optical tracking at **60 Hz**; RGB/RGB-D/US at **30 FPS**. +- **Delay measurement:** the robot executes controlled sinusoidal motions while each sensor stream observes the motion (needle/probe/camera movement). Phase differences between each sensor’s measured sinusoid and the robot reference are used to estimate per-stream delays. The delay was calculated using [PLUS Toolkit](https://pmc.ncbi.nlm.nih.gov/articles/PMC4437531/). +- **Alignment:** timestamps are corrected using measured delays, and asynchronous streams are aligned via interpolation/resampling to produce unified, time-aligned trajectories. +- **Recording:** all raw streams and teleop inputs are recorded in ROS (rosbag). + +--- + +## πŸ‘₯ Attribution & Contact + +| | | +|:---------------------------|:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Dataset Lead** | Shing Shin Cheng; Wei Wang; Qingpeng Ding; Yuelin Zhang; Zhouyang Hong; Luoyao Kang; Wenxuan Xie | +| **Institution** | The Chinese University of Hong Kong (CUHK); First Affiliated Hospital, Sun Yat-Sen University | +| **Contact Email** | sscheng@cuhk.edu.hk | +| **Citation (BibTeX)** | 
@misc{openh_ausnb_2026,
  author = {Cheng, Shing Shin and Wang, Wei and Ding, Qingpeng and Zhang, Yuelin and Hong, Zhouyang and Kang, Luoyao and Xie, Wenxuan},
  title = {A Multimodal Dataset for Autonomous Probe Placement and Needle Retrieval in Ultrasound-Guided Liver Biopsy (US-PPNR)},
  year = {2026},
  publisher = {Open-H-Embodiment},
  license = {CC BY 4.0}, 
}
| +--- \ No newline at end of file diff --git a/Ultrasound/cuhk/hongliang_ren/tracked_eus/README.md b/Ultrasound/cuhk/hongliang_ren/tracked_eus/README.md new file mode 100644 index 0000000000000000000000000000000000000000..77cd5a4c461a2dad160942461f5501021aad8fab --- /dev/null +++ b/Ultrasound/cuhk/hongliang_ren/tracked_eus/README.md @@ -0,0 +1,115 @@ + + +# [tracked_EUS] - README + +--- + +## πŸ“‹ At a Glance + +* Endoscopic ultrasound (EUS) scanning with 6-DoF pose tracking on in-vivo porcine models.* + +--- + +## πŸ“– Dataset Overview + +*This dataset consists of 34 trajectories of expert gastroenterologist performing EUS examinations on in-vivo porcine models +The data contains synchronized data streams including B-mode ultrasound frames, white-light endoscopic video, +and 6-DoF pose ground truth recorded via electromagnetic (EM) tracking, +capturing standard clinical scanning procedures across the upper and lower digestive tracts.* + +| | | +| :--- |:--------------------------------------------------------------------------------------------------------------------------------------------------| +| **Total Trajectories** | `[34]` | +| **Data Type** | `[ ] Clinical` `[ ] Ex-Vivo` `[ ] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[x] In-Vivo (Porcine Model)` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [x] **Endoscopic ultrasound** + +### Demonstrated Skills + +The dataset demonstrates expert maneuvering and scanning of specific anatomical stations: + +* **Vascular Scanning**: Inferior Vena Cava (IVC), Portal Vein (PV), Thoracic Aorta. +* **Organ Scanning**: Pancreas, Liver, Mediastinum. +* **Station Navigation**: Gastric station scanning, Gastric Cardia entry. +* **Procedure Phases**: Endoscope insertion, Transrectal EUS survey. + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [x] **Human Operation** + +### Operator Details + +| | Description | +| :--- |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[x] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-11-1]` to `[2026-1-30]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** +--- + +## 🎯 Action & State Space Representation + +### State Space +- **`observation.state`**: `[tx, ty, tz, qx, qy, qz, qw, quality]` + - EM Tracker pose of the probe tip relative to the emitter base. + - Index 7 represents the tracker signal quality. + +### Action Space +- **`action`**: `[tx, ty, tz, qx, qy, qz, qw]` + - Target pose for the next timestep (Delta-time prediction target). + +### Metadata Fields +- **`observation.meta.probe_acquisition_param`**: `[freq, depth, gain, fps]` +- **`observation.meta.prob_cali_mtx`**: **Placeholder (Identity)**. + - *Note: Calibration between the Ultrasound image plane and the EM sensor is not currently provided. This field contains identity transforms for structural compliance.* + + +--- + +## ⏱️ Data Synchronization Approach + +*Temporal synchronization was achieved through a motion-based calibration involving periodic Z-axis probe movements +in a water tank. We calculated the optimal time offset by cross-correlating the images intensity waveform +with the EM sensor's data.* + +--- + +## πŸ‘₯ Attribution & Contact + +| | | +| :--- |:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Dataset Lead** | `[Erqi Wang, Rui Ji, Zhen Li, Ning Zhong, Xiuli Zuo, Hongliang Ren]` | +| **Institution** | `[Chinese University of Hong Kong, Qilu Hospital of Shandong University]` | +| **Contact Email** | `[ewang@link.cuhk.edu.hk]` | +| **Citation (BibTeX)** | 
@misc{[Tracked_EUS],
  author = {[Erqi Wang, Rui Ji, Zhen Li, Ning Zhong, Xiuli Zuo, Hongliang Ren]},
  title = {[Tracked_EUS]},
  year = {2025},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Ultrasound/cuhk/hongliang_ren/tracked_us/README.md b/Ultrasound/cuhk/hongliang_ren/tracked_us/README.md new file mode 100644 index 0000000000000000000000000000000000000000..3f3a3c7a77fe6fd567ff34eefa79a147d655e962 --- /dev/null +++ b/Ultrasound/cuhk/hongliang_ren/tracked_us/README.md @@ -0,0 +1,115 @@ + + +# [tracked_US] - README + +--- + +## πŸ“‹ At a Glance + +* Freehand ultrasound scanning on human forearm with 6-DoF pose tracking.* + +--- + +## πŸ“– Dataset Overview + +*This dataset consists of 34 trajectories of expert gastroenterologist performing EUS examinations on in-vivo porcine models +The data contains synchronized data streams including B-mode ultrasound frames, white-light endoscopic video, +and 6-DoF pose ground truth recorded via electromagnetic (EM) tracking, +capturing standard clinical scanning procedures across the upper and lower digestive tracts.* + +| | | +| :--- |:-------------------------------------------------------------------------------------------------------------------------------| +| **Total Trajectories** | `[30]` | +| **Data Type** | `[x] Clinical (Human Volunteer)` `[ ] Ex-Vivo` `[ ] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` | +| **License** | CC BY 4.0 | +| **Version** | `[e.g., 1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [ ] **Ultrasound Robotics** +- [x] **Endoscopic ultrasound** + +### Demonstrated Skills + +The dataset demonstrates systematic freehand scanning patterns used for sensor calibration and surface anatomical mapping: + +* **Linear Scanning**: Straight-line sweeps along the forearm (parallel/perpendicular). +* **Curvilinear Scanning**: C-shape trajectories manipulating probe orientation. +* **Complex Trajectories**: S-shape scanning covering larger surface areas. +* **Probe Orientation Control**: Maintaining contact while varying orientation (Parallel vs. Perpendicular to arm axis). + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [ ] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [x] **Human Operation** + +### Operator Details + +| | Description | +| :--- |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[x] Expert (e.g., Surgeon, Sonographer)`
`[ ] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-11-1]` to `[2026-1-30]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** +--- + +## 🎯 Action & State Space Representation + +### State Space +- **`observation.state`**: `[tx, ty, tz, qx, qy, qz, qw, quality]` + - EM Tracker pose of the probe tip relative to the emitter base. + - Index 7 represents the tracker signal quality. + +### Action Space +- **`action`**: `[tx, ty, tz, qx, qy, qz, qw]` + - Target pose for the next timestep (Delta-time prediction target). + +### Metadata Fields +- **`observation.meta.probe_acquisition_param`**: `[freq, depth, gain, fps]` +- **`observation.meta.prob_cali_mtx`**: **Placeholder (Identity)**. + - *Note: Calibration between the Ultrasound image plane and the EM sensor is not currently provided. This field contains identity transforms for structural compliance.* + + +--- + +## ⏱️ Data Synchronization Approach + +*Temporal synchronization was achieved through a motion-based calibration involving periodic Z-axis probe movements +in a water tank. We calculated the optimal time offset by cross-correlating the images intensity waveform +with the EM sensor's data.* + +--- + +## πŸ‘₯ Attribution & Contact + +| | | +| :--- |:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **Dataset Lead** | `[Erqi Wang, Rui Ji, Zhen Li, Ning Zhong, Xiuli Zuo, Hongliang Ren]` | +| **Institution** | `[Chinese University of Hong Kong, Qilu Hospital of Shandong University]` | +| **Contact Email** | `[ewang@link.cuhk.edu.hk]` | +| **Citation (BibTeX)** | 
@misc{[Tracked_US],
  author = {[Erqi Wang, Rui Ji, Zhen Li, Ning Zhong, Xiuli Zuo, Hongliang Ren]},
  title = {[Tracked_US]},
  year = {2025},
  publisher = {Open-H-Embodiment},
  note = {https://hrpp.research.virginia.edu/teams/irb-sbs/researcher-guide-irb-sbs/identifiers}
}
| diff --git a/Ultrasound/cuhk/xiangyu_chu/bmt_insertion_dataset/bmt_insertion_dataset/README.md b/Ultrasound/cuhk/xiangyu_chu/bmt_insertion_dataset/bmt_insertion_dataset/README.md new file mode 100644 index 0000000000000000000000000000000000000000..e140051b5d388309dbe66b9f7d89a746fc49ad29 --- /dev/null +++ b/Ultrasound/cuhk/xiangyu_chu/bmt_insertion_dataset/bmt_insertion_dataset/README.md @@ -0,0 +1,234 @@ + + +# [Dataset Name] - README + +--- + +## πŸ“‹ At a Glance + +Static dual-arm in-plane needle insertion with interferences + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 2000 trajectories of a trained researcher using the dual-arm UR5e to perform needle insertion tasks on ex-vivo tissues include objects: mimicked vessel, different simulated tumors (grape, blueberry and fish ball), pork lean, bone ,liver, kidney.* + + +| | | +| :--- | :--- | +| **Total Trajectories** | `[2000]` | +| **Total Hours** | `[25]` | +| **Data Type** | `[ ] Clinical` `[x] Ex-Vivo` `[ ] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*Autonomous dual-arm robotic needle insertion.* + + + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [x] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[x] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-12-26]` to `[2026-01-07]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + + + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [ ] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +After each demonstration, we adjust the initial position of at least one of the robotic arm, ultrasonic probe, and target object (phantom). + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `[UR5e]` +- **Robot 2:** `[UR5e]` + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Room/3rd Person Camera** | `[Intel RealSense D435i @ 30fps]` | +| **Medical Imager** | `[Siemens ACUSON Juniper Ultrasound System with 12L3 Transducer, B-Mode]` | + + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [x] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [x] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + + + + +``` +action: [j1, j2, j3, j4, j5, j6, j7, j8, j9, j10, j11, j12] +- 'L_Base', 'L_Shoulder', 'L_Elbow', 'L_Wrist1', 'L_Wrist2' 'L_Wrist3' 'R_Base', 'R_Shoulder', 'R_Elbow', 'R_Wrist1', 'R_Wrist2', 'R_Wrist3' : Absolute joint positions for two 6-DOF arms (radians) +action.left: [j1, j2, j3, j4, j5, j6] +- 'L_Base', 'L_Shoulder', 'L_Elbow', 'L_Wrist1', 'L_Wrist2' 'L_Wrist3' : Absolute joint positions for 6-DOF arm (radians) +action.right: [j1, j2, j3, j4, j5, j6] +- 'R_Base', 'R_Shoulder', 'R_Elbow', 'R_Wrist1', 'R_Wrist2', 'R_Wrist3' : Absolute joint positions for 6-DOF arm (radians) +``` + + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + + + +**Example:** +``` +observation.state: ["left_x", "left_y", "left_z", "left_qx", "left_qy", "left_qz", "left_qw", "right_x", "right_y", "right_z", "right_qx", "right_qy", "right_qz", "right_qw"] +- left_x, left_y, left_z, right_x, right_y, right_z: Absolute position in left and right robot base frame (meters) +- left_qx, left_qy, left_qz, left_qw, right_qx, "right_qy, right_qz, right_qw: Absolute orientation as quaternion +observation.state.left: [x, y, z, qx, qy, qz, qw] +- x, y, z: Absolute position in left robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +observation.state.right: [x, y, z, qx, qy, qz, qw] +- x, y, z: Absolute position in right robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +``` + +### πŸ“‹ Recommended Additional Representations +**Step Information Included:** +- [x] **Step Labels** + +**Example:** +``` +observation.step_label: ["label"] +- label: step labels represent the robot motion state, for each trajectory includes 3 labels and 5 motion sequences ('hold - alignment - hold - insertion - hold'). +``` + + + +--- + +## ⏱️ Data Synchronization Approach + + +*We collect joint kinematics from our UR5e, RGB-D frames from Intel RealSense D435 cameras and US images from Siemens Acuson Juniper with 12L3 transducer, all running in ROS 1 Noetic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /frame_grabber/us_img, /camera_top/color/image_raw, /right_tcp_pos, and /left_tcp_pos in a single rosbag session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±100 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Zhongyu Chen, Yunxi Tang, Tianqi Yang, Xiangyu Chu]` | +| **Institution** | `[Bioinspired Robotics and Medical Technology (BMT) Lab, Department of Mechanical and Automation Engineering (MAE), The Chinese University of Hong Kong (CUHK) ]` | +| **Contact Email** | `[zhongyuchen001@cuhk.edu.hk, 1155245457@link.cuhk.edu.hk, tianqiyang@cuhk.edu.hk, xychu@mae.cuhk.edu.hk]` | + \ No newline at end of file diff --git a/Ultrasound/cuhk/xiangyu_chu/bmt_insertion_tumor_fishball_dataset/bmt_insertion_tumor_fishball_dataset/README.md b/Ultrasound/cuhk/xiangyu_chu/bmt_insertion_tumor_fishball_dataset/bmt_insertion_tumor_fishball_dataset/README.md new file mode 100644 index 0000000000000000000000000000000000000000..3e5903f6b438c1d14122155feb8dcf24fe5b1919 --- /dev/null +++ b/Ultrasound/cuhk/xiangyu_chu/bmt_insertion_tumor_fishball_dataset/bmt_insertion_tumor_fishball_dataset/README.md @@ -0,0 +1,234 @@ + + +# [Dataset Name] - README + +--- + +## πŸ“‹ At a Glance + +Static dual-arm in-plane needle insertion with interferences + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 2000 trajectories of a trained researcher using the dual-arm UR5e to perform needle insertion tasks on ex-vivo tissues include objects: mimicked vessel, different simulated tumors (grape, blueberry and fish ball), pork lean, bone ,liver, kidney.* + + +| | | +| :--- | :--- | +| **Total Trajectories** | `[2000]` | +| **Total Hours** | `[25]` | +| **Data Type** | `[ ] Clinical` `[x] Ex-Vivo` `[ ] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*Autonomous dual-arm robotic needle insertion.* + + + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [x] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[x] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-12-26]` to `[2026-01-07]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + + + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [ ] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +After each demonstration, we adjust the initial position of at least one of the robotic arm, ultrasonic probe, and target object (phantom). + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `[UR5e]` +- **Robot 2:** `[UR5e]` + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Room/3rd Person Camera** | `[Intel RealSense D435i @ 30fps]` | +| **Medical Imager** | `[Siemens ACUSON Juniper Ultrasound System with 12L3 Transducer, B-Mode]` | + + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + + + + +``` +action: ["left_x", "left_y", "left_z", "left_qx", "left_qy", "left_qz", "left_qw", "right_x", "right_y", "right_z", "right_qx", "right_qy", "right_qz", "right_qw"] +- left_x, left_y, left_z, right_x, right_y, right_z: Absolute position in left and right robot base frame (meters) +- left_qx, left_qy, left_qz, left_qw, right_qx, "right_qy, right_qz, right_qw: Absolute orientation as quaternion +action.left: [x, y, z, qx, qy, qz, qw] +- x, y, z: Absolute position in left robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +action.right: [x, y, z, qx, qy, qz, qw] +- x, y, z: Absolute position in right robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +``` + + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [ ] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + + + +**Example:** +``` +observation.state: [j1, j2, j3, j4, j5, j6, j7, j8, j9, j10, j11, j12] +- 'L_Base', 'L_Shoulder', 'L_Elbow', 'L_Wrist1', 'L_Wrist2' 'L_Wrist3' 'R_Base', 'R_Shoulder', 'R_Elbow', 'R_Wrist1', 'R_Wrist2', 'R_Wrist3' : Absolute joint positions for two 6-DOF arms (radians) +observation.state.left: [j1, j2, j3, j4, j5, j6] +- 'L_Base', 'L_Shoulder', 'L_Elbow', 'L_Wrist1', 'L_Wrist2' 'L_Wrist3' : Absolute joint positions for 6-DOF arm (radians) +observation.state.right: [j1, j2, j3, j4, j5, j6] +- 'R_Base', 'R_Shoulder', 'R_Elbow', 'R_Wrist1', 'R_Wrist2', 'R_Wrist3' : Absolute joint positions for 6-DOF arm (radians) +``` + +### πŸ“‹ Recommended Additional Representations +**Step Information Included:** +- [x] **Step Labels** + +**Example:** +``` +observation.step_label: ["label"] +- label: step labels represent the robot motion state, for each trajectory includes 3 labels and 5 motion sequences ('hold - alignment - hold - insertion - hold'). +``` + + + +--- + +## ⏱️ Data Synchronization Approach + + +*We collect joint kinematics from our UR5e, RGB-D frames from Intel RealSense D435 cameras and US images from Siemens Acuson Juniper with 12L3 transducer, all running in ROS 1 Noetic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /frame_grabber/us_img, /camera_top/color/image_raw, /right_tcp_pos, and /left_tcp_pos in a single rosbag session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±100 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Zhongyu Chen, Yunxi Tang, Tianqi Yang, Xiangyu Chu]` | +| **Institution** | `[Bioinspired Robotics and Medical Technology (BMT) Lab, Department of Mechanical and Automation Engineering (MAE), The Chinese University of Hong Kong (CUHK) ]` | +| **Contact Email** | `[zhongyuchen001@cuhk.edu.hk, 1155245457@link.cuhk.edu.hk, tianqiyang@cuhk.edu.hk, xychu@mae.cuhk.edu.hk]` | + \ No newline at end of file diff --git a/Ultrasound/cuhk/xiangyu_chu/bmt_insertion_vessel_dataset/bmt_insertion_vessel_dataset/README.md b/Ultrasound/cuhk/xiangyu_chu/bmt_insertion_vessel_dataset/bmt_insertion_vessel_dataset/README.md new file mode 100644 index 0000000000000000000000000000000000000000..3e5903f6b438c1d14122155feb8dcf24fe5b1919 --- /dev/null +++ b/Ultrasound/cuhk/xiangyu_chu/bmt_insertion_vessel_dataset/bmt_insertion_vessel_dataset/README.md @@ -0,0 +1,234 @@ + + +# [Dataset Name] - README + +--- + +## πŸ“‹ At a Glance + +Static dual-arm in-plane needle insertion with interferences + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 2000 trajectories of a trained researcher using the dual-arm UR5e to perform needle insertion tasks on ex-vivo tissues include objects: mimicked vessel, different simulated tumors (grape, blueberry and fish ball), pork lean, bone ,liver, kidney.* + + +| | | +| :--- | :--- | +| **Total Trajectories** | `[2000]` | +| **Total Hours** | `[25]` | +| **Data Type** | `[ ] Clinical` `[x] Ex-Vivo` `[ ] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*Autonomous dual-arm robotic needle insertion.* + + + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [x] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[x] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-12-26]` to `[2026-01-07]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + + + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [ ] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +After each demonstration, we adjust the initial position of at least one of the robotic arm, ultrasonic probe, and target object (phantom). + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `[UR5e]` +- **Robot 2:** `[UR5e]` + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Room/3rd Person Camera** | `[Intel RealSense D435i @ 30fps]` | +| **Medical Imager** | `[Siemens ACUSON Juniper Ultrasound System with 12L3 Transducer, B-Mode]` | + + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + + + + +``` +action: ["left_x", "left_y", "left_z", "left_qx", "left_qy", "left_qz", "left_qw", "right_x", "right_y", "right_z", "right_qx", "right_qy", "right_qz", "right_qw"] +- left_x, left_y, left_z, right_x, right_y, right_z: Absolute position in left and right robot base frame (meters) +- left_qx, left_qy, left_qz, left_qw, right_qx, "right_qy, right_qz, right_qw: Absolute orientation as quaternion +action.left: [x, y, z, qx, qy, qz, qw] +- x, y, z: Absolute position in left robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +action.right: [x, y, z, qx, qy, qz, qw] +- x, y, z: Absolute position in right robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +``` + + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [ ] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + + + +**Example:** +``` +observation.state: [j1, j2, j3, j4, j5, j6, j7, j8, j9, j10, j11, j12] +- 'L_Base', 'L_Shoulder', 'L_Elbow', 'L_Wrist1', 'L_Wrist2' 'L_Wrist3' 'R_Base', 'R_Shoulder', 'R_Elbow', 'R_Wrist1', 'R_Wrist2', 'R_Wrist3' : Absolute joint positions for two 6-DOF arms (radians) +observation.state.left: [j1, j2, j3, j4, j5, j6] +- 'L_Base', 'L_Shoulder', 'L_Elbow', 'L_Wrist1', 'L_Wrist2' 'L_Wrist3' : Absolute joint positions for 6-DOF arm (radians) +observation.state.right: [j1, j2, j3, j4, j5, j6] +- 'R_Base', 'R_Shoulder', 'R_Elbow', 'R_Wrist1', 'R_Wrist2', 'R_Wrist3' : Absolute joint positions for 6-DOF arm (radians) +``` + +### πŸ“‹ Recommended Additional Representations +**Step Information Included:** +- [x] **Step Labels** + +**Example:** +``` +observation.step_label: ["label"] +- label: step labels represent the robot motion state, for each trajectory includes 3 labels and 5 motion sequences ('hold - alignment - hold - insertion - hold'). +``` + + + +--- + +## ⏱️ Data Synchronization Approach + + +*We collect joint kinematics from our UR5e, RGB-D frames from Intel RealSense D435 cameras and US images from Siemens Acuson Juniper with 12L3 transducer, all running in ROS 1 Noetic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /frame_grabber/us_img, /camera_top/color/image_raw, /right_tcp_pos, and /left_tcp_pos in a single rosbag session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±100 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Zhongyu Chen, Yunxi Tang, Tianqi Yang, Xiangyu Chu]` | +| **Institution** | `[Bioinspired Robotics and Medical Technology (BMT) Lab, Department of Mechanical and Automation Engineering (MAE), The Chinese University of Hong Kong (CUHK) ]` | +| **Contact Email** | `[zhongyuchen001@cuhk.edu.hk, 1155245457@link.cuhk.edu.hk, tianqiyang@cuhk.edu.hk, xychu@mae.cuhk.edu.hk]` | + \ No newline at end of file diff --git a/Ultrasound/cuhk/xiangyu_chu/bmt_needle_insertion_dataset/README.md b/Ultrasound/cuhk/xiangyu_chu/bmt_needle_insertion_dataset/README.md new file mode 100644 index 0000000000000000000000000000000000000000..e140051b5d388309dbe66b9f7d89a746fc49ad29 --- /dev/null +++ b/Ultrasound/cuhk/xiangyu_chu/bmt_needle_insertion_dataset/README.md @@ -0,0 +1,234 @@ + + +# [Dataset Name] - README + +--- + +## πŸ“‹ At a Glance + +Static dual-arm in-plane needle insertion with interferences + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 2000 trajectories of a trained researcher using the dual-arm UR5e to perform needle insertion tasks on ex-vivo tissues include objects: mimicked vessel, different simulated tumors (grape, blueberry and fish ball), pork lean, bone ,liver, kidney.* + + +| | | +| :--- | :--- | +| **Total Trajectories** | `[2000]` | +| **Total Hours** | `[25]` | +| **Data Type** | `[ ] Clinical` `[x] Ex-Vivo` `[ ] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*Autonomous dual-arm robotic needle insertion.* + + + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [x] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[x] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-12-26]` to `[2026-01-07]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + + + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [ ] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +After each demonstration, we adjust the initial position of at least one of the robotic arm, ultrasonic probe, and target object (phantom). + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `[UR5e]` +- **Robot 2:** `[UR5e]` + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Room/3rd Person Camera** | `[Intel RealSense D435i @ 30fps]` | +| **Medical Imager** | `[Siemens ACUSON Juniper Ultrasound System with 12L3 Transducer, B-Mode]` | + + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [x] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [x] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + + + + +``` +action: [j1, j2, j3, j4, j5, j6, j7, j8, j9, j10, j11, j12] +- 'L_Base', 'L_Shoulder', 'L_Elbow', 'L_Wrist1', 'L_Wrist2' 'L_Wrist3' 'R_Base', 'R_Shoulder', 'R_Elbow', 'R_Wrist1', 'R_Wrist2', 'R_Wrist3' : Absolute joint positions for two 6-DOF arms (radians) +action.left: [j1, j2, j3, j4, j5, j6] +- 'L_Base', 'L_Shoulder', 'L_Elbow', 'L_Wrist1', 'L_Wrist2' 'L_Wrist3' : Absolute joint positions for 6-DOF arm (radians) +action.right: [j1, j2, j3, j4, j5, j6] +- 'R_Base', 'R_Shoulder', 'R_Elbow', 'R_Wrist1', 'R_Wrist2', 'R_Wrist3' : Absolute joint positions for 6-DOF arm (radians) +``` + + +### State Space Representation + +**State Information Included:** +- [ ] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [x] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + + + +**Example:** +``` +observation.state: ["left_x", "left_y", "left_z", "left_qx", "left_qy", "left_qz", "left_qw", "right_x", "right_y", "right_z", "right_qx", "right_qy", "right_qz", "right_qw"] +- left_x, left_y, left_z, right_x, right_y, right_z: Absolute position in left and right robot base frame (meters) +- left_qx, left_qy, left_qz, left_qw, right_qx, "right_qy, right_qz, right_qw: Absolute orientation as quaternion +observation.state.left: [x, y, z, qx, qy, qz, qw] +- x, y, z: Absolute position in left robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +observation.state.right: [x, y, z, qx, qy, qz, qw] +- x, y, z: Absolute position in right robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +``` + +### πŸ“‹ Recommended Additional Representations +**Step Information Included:** +- [x] **Step Labels** + +**Example:** +``` +observation.step_label: ["label"] +- label: step labels represent the robot motion state, for each trajectory includes 3 labels and 5 motion sequences ('hold - alignment - hold - insertion - hold'). +``` + + + +--- + +## ⏱️ Data Synchronization Approach + + +*We collect joint kinematics from our UR5e, RGB-D frames from Intel RealSense D435 cameras and US images from Siemens Acuson Juniper with 12L3 transducer, all running in ROS 1 Noetic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /frame_grabber/us_img, /camera_top/color/image_raw, /right_tcp_pos, and /left_tcp_pos in a single rosbag session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±100 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Zhongyu Chen, Yunxi Tang, Tianqi Yang, Xiangyu Chu]` | +| **Institution** | `[Bioinspired Robotics and Medical Technology (BMT) Lab, Department of Mechanical and Automation Engineering (MAE), The Chinese University of Hong Kong (CUHK) ]` | +| **Contact Email** | `[zhongyuchen001@cuhk.edu.hk, 1155245457@link.cuhk.edu.hk, tianqiyang@cuhk.edu.hk, xychu@mae.cuhk.edu.hk]` | + \ No newline at end of file diff --git a/Ultrasound/cuhk/xiangyu_chu/bmt_tumor_grape_insertion_dataset/bmt_tumor_grape_insertion_dataset/README.md b/Ultrasound/cuhk/xiangyu_chu/bmt_tumor_grape_insertion_dataset/bmt_tumor_grape_insertion_dataset/README.md new file mode 100644 index 0000000000000000000000000000000000000000..3e5903f6b438c1d14122155feb8dcf24fe5b1919 --- /dev/null +++ b/Ultrasound/cuhk/xiangyu_chu/bmt_tumor_grape_insertion_dataset/bmt_tumor_grape_insertion_dataset/README.md @@ -0,0 +1,234 @@ + + +# [Dataset Name] - README + +--- + +## πŸ“‹ At a Glance + +Static dual-arm in-plane needle insertion with interferences + +--- + +## πŸ“– Dataset Overview + +*This dataset contains 2000 trajectories of a trained researcher using the dual-arm UR5e to perform needle insertion tasks on ex-vivo tissues include objects: mimicked vessel, different simulated tumors (grape, blueberry and fish ball), pork lean, bone ,liver, kidney.* + + +| | | +| :--- | :--- | +| **Total Trajectories** | `[2000]` | +| **Total Hours** | `[25]` | +| **Data Type** | `[ ] Clinical` `[x] Ex-Vivo` `[ ] Table-Top Phantom` `[ ] Digital Simulation` `[ ] Physical Simulation` `[ ] Other (If checked, update "Other")` | +| **License** | CC BY 4.0 | +| **Version** | `[1.0]` | + +--- + +## 🎯 Tasks & Domain + +### Domain + +*Select the primary domain for this dataset.* + +- [ ] **Surgical Robotics** +- [x] **Ultrasound Robotics** +- [ ] **Other Healthcare Robotics** (Please specify: `[Your Domain]`) + +### Demonstrated Skills + +*Autonomous dual-arm robotic needle insertion.* + + + +--- + +## πŸ”¬ Data Collection Details + +### Collection Method + +*How was the data collected?* + +- [ ] **Human Teleoperation** +- [x] **Programmatic/State-Machine** +- [ ] **AI Policy / Autonomous** +- [ ] **Other** (Please specify: `[Your Method]`) + +### Operator Details + +| | Description | +| :--- | :--- | +| **Operator Count** | `[1]` | +| **Operator Skill Level** | `[ ] Expert (e.g., Surgeon, Sonographer)`
`[x] Intermediate (e.g., Trained Researcher)`
`[ ] Novice (e.g., ML Researcher with minimal experience)`
`[ ] N/A` | +| **Collection Period** | From `[2025-12-26]` to `[2026-01-07]` | + +### Recovery Demonstrations + +*Does this dataset include examples of recovering from failure?* + +- [ ] **Yes** +- [x] **No** + + + +--- + +## πŸ’‘ Diversity Dimensions + +*Check all dimensions that were intentionally varied during data collection.* + +- [ ] **Camera Position / Angle** +- [ ] **Lighting Conditions** +- [x] **Target Object** (e.g., different phantom models, suture types) +- [x] **Spatial Layout** (e.g., placing the target suture needle in various locations) +- [ ] **Robot Embodiment** (if multiple robots were used) +- [ ] **Task Execution** (e.g., different techniques for the same task) +- [ ] **Background / Scene** +- [ ] **Other** (Please specify: `[Your Dimension]`) + +*If you checked any of the above please briefly elaborate below.* + +After each demonstration, we adjust the initial position of at least one of the robotic arm, ultrasonic probe, and target object (phantom). + +--- + +## πŸ› οΈ Equipment & Setup + +### Robotic Platform(s) + +*List the primary robot(s) used.* + +- **Robot 1:** `[UR5e]` +- **Robot 2:** `[UR5e]` + +### Sensors & Cameras + +*List the sensors and cameras used. Specify model names where possible. (Add and remove rows as needed)* + +| Type | Model/Details | +| :--- | :--- | +| **Room/3rd Person Camera** | `[Intel RealSense D435i @ 30fps]` | +| **Medical Imager** | `[Siemens ACUSON Juniper Ultrasound System with 12L3 Transducer, B-Mode]` | + + +--- + +## 🎯 Action & State Space Representation + +*Describe how actions and robot states are represented in your dataset. This is crucial for understanding data compatibility and enabling effective policy learning.* + +### Action Space Representation + +**Primary Action Representation:** +- [x] **Absolute Cartesian** (position/orientation relative to robot base) +- [ ] **Relative Cartesian** (delta position/orientation from current pose) +- [ ] **Joint Space** (direct joint angle commands) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Orientation Representation:** +- [x] **Quaternions** (x, y, z, w) +- [ ] **Euler Angles** (roll, pitch, yaw) +- [ ] **Axis-Angle** (rotation vector) +- [ ] **Rotation Matrix** (3x3 matrix) +- [ ] **Other** (Please specify: `[Your Representation]`) + +**Reference Frame:** +- [x] **Robot Base Frame** +- [ ] **Tool/End-Effector Frame** +- [ ] **World/Global Frame** +- [ ] **Camera Frame** +- [ ] **Other** (Please specify: `[Your Frame]`) + + + + +``` +action: ["left_x", "left_y", "left_z", "left_qx", "left_qy", "left_qz", "left_qw", "right_x", "right_y", "right_z", "right_qx", "right_qy", "right_qz", "right_qw"] +- left_x, left_y, left_z, right_x, right_y, right_z: Absolute position in left and right robot base frame (meters) +- left_qx, left_qy, left_qz, left_qw, right_qx, "right_qy, right_qz, right_qw: Absolute orientation as quaternion +action.left: [x, y, z, qx, qy, qz, qw] +- x, y, z: Absolute position in left robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +action.right: [x, y, z, qx, qy, qz, qw] +- x, y, z: Absolute position in right robot base frame (meters) +- qx, qy, qz, qw: Absolute orientation as quaternion +``` + + +### State Space Representation + +**State Information Included:** +- [x] **Joint Positions** (all articulated joints) +- [ ] **Joint Velocities** +- [ ] **End-Effector Pose** (Cartesian position/orientation) +- [ ] **Force/Torque Readings** +- [ ] **Gripper State** (position, force, etc.) +- [ ] **Other** (Please specify: `[Your State Info]`) + + + +**Example:** +``` +observation.state: [j1, j2, j3, j4, j5, j6, j7, j8, j9, j10, j11, j12] +- 'L_Base', 'L_Shoulder', 'L_Elbow', 'L_Wrist1', 'L_Wrist2' 'L_Wrist3' 'R_Base', 'R_Shoulder', 'R_Elbow', 'R_Wrist1', 'R_Wrist2', 'R_Wrist3' : Absolute joint positions for two 6-DOF arms (radians) +observation.state.left: [j1, j2, j3, j4, j5, j6] +- 'L_Base', 'L_Shoulder', 'L_Elbow', 'L_Wrist1', 'L_Wrist2' 'L_Wrist3' : Absolute joint positions for 6-DOF arm (radians) +observation.state.right: [j1, j2, j3, j4, j5, j6] +- 'R_Base', 'R_Shoulder', 'R_Elbow', 'R_Wrist1', 'R_Wrist2', 'R_Wrist3' : Absolute joint positions for 6-DOF arm (radians) +``` + +### πŸ“‹ Recommended Additional Representations +**Step Information Included:** +- [x] **Step Labels** + +**Example:** +``` +observation.step_label: ["label"] +- label: step labels represent the robot motion state, for each trajectory includes 3 labels and 5 motion sequences ('hold - alignment - hold - insertion - hold'). +``` + + + +--- + +## ⏱️ Data Synchronization Approach + + +*We collect joint kinematics from our UR5e, RGB-D frames from Intel RealSense D435 cameras and US images from Siemens Acuson Juniper with 12L3 transducer, all running in ROS 1 Noetic on the same workstation clocked with ROS Time. Both drivers stamp their outgoing messages’ header.stamp fields with the shared system clock, and we record /joint_states, /frame_grabber/us_img, /camera_top/color/image_raw, /right_tcp_pos, and /left_tcp_pos in a single rosbag session. During export to LeRobot, each data point’s ROS header.stamp is written verbatim into the timestamp attribute. Offline checks show inter-sensor skew stays below Β±100 ms across a 2-minute capture.* + +--- + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `[Zhongyu Chen, Yunxi Tang, Tianqi Yang, Xiangyu Chu]` | +| **Institution** | `[Bioinspired Robotics and Medical Technology (BMT) Lab, Department of Mechanical and Automation Engineering (MAE), The Chinese University of Hong Kong (CUHK) ]` | +| **Contact Email** | `[zhongyuchen001@cuhk.edu.hk, 1155245457@link.cuhk.edu.hk, tianqiyang@cuhk.edu.hk, xychu@mae.cuhk.edu.hk]` | + \ No newline at end of file diff --git a/Ultrasound/hkbu/ultravision_lab/3Dircadb1.1/README.md b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.1/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.1/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/3Dircadb1.10/README.md b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.10/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.10/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/3Dircadb1.11/README.md b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.11/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.11/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/3Dircadb1.12/README.md b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.12/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.12/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/3Dircadb1.13/README.md b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.13/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.13/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/3Dircadb1.14/README.md b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.14/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.14/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/3Dircadb1.15/README.md b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.15/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.15/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/3Dircadb1.16/README.md b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.16/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.16/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/3Dircadb1.17/README.md b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.17/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.17/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/3Dircadb1.20/README.md b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.20/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.20/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/3Dircadb1.4/README.md b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.4/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.4/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/3Dircadb1.5/README.md b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.5/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.5/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/3Dircadb1.6/README.md b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.6/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.6/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/3Dircadb1.8/README.md b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.8/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.8/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/3Dircadb1.9/README.md b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.9/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/3Dircadb1.9/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1001/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1001/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1001/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1004/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1004/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1004/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1006/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1006/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1006/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1007/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1007/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1007/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1009/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1009/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1009/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1010/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1010/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1010/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1012/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1012/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1012/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1013/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1013/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1013/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1014/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1014/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1014/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1017/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1017/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1017/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1022/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1022/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1022/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1023/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1023/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1023/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1025/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1025/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1025/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1027/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1027/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1027/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1031/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1031/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1031/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1033/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1033/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1033/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1035/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1035/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1035/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1038/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1038/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1038/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1039/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1039/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1039/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1041/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1041/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1041/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1042/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1042/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1042/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1043/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1043/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1043/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1045/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1045/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1045/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1046/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1046/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1046/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1049/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1049/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1049/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1050/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1050/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1050/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1051/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1051/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1051/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1053/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1053/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1053/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1054/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1054/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1054/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1056/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1056/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1056/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1058/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1058/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1058/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1060/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1060/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1060/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1062/README.md b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1062/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/CRLM-CT-1062/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/hp001/README.md b/Ultrasound/hkbu/ultravision_lab/hp001/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/hp001/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/hp004/README.md b/Ultrasound/hkbu/ultravision_lab/hp004/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/hp004/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/hkbu/ultravision_lab/hp005/README.md b/Ultrasound/hkbu/ultravision_lab/hp005/README.md new file mode 100644 index 0000000000000000000000000000000000000000..b75a53e58452c53b5b1e8628ef5ff8843f0002c7 --- /dev/null +++ b/Ultrasound/hkbu/ultravision_lab/hp005/README.md @@ -0,0 +1,90 @@ +# Ultrasound Probe Guidance LeRobot Dataset + +## Overview + +This dataset contains synthetic ultrasound probe guidance data in LeRobot v2.1 format for Vision-Language-Action (VLA) model training. The data simulates ultrasound imaging of abdominal organs (primarily liver) with probe trajectories along skin surfaces. + +## Data Structure + +``` +lerobot_dataset/ +β”œβ”€β”€ videos/ +β”‚ └── episode_XXXXXX/ +β”‚ └── observation.images.ultrasound.mp4 +β”œβ”€β”€ data/ +β”‚ └── chunk-000/ +β”‚ └── episode_XXXXXX.parquet +└── meta/ + β”œβ”€β”€ info.json + β”œβ”€β”€ episodes.jsonl + β”œβ”€β”€ episodes_stats.jsonl + β”œβ”€β”€ tasks.jsonl + └── README.md +``` + +## Path Segment Definitions (seg_1 to seg_12) + +### Standard Planes + +| Plane | Name | Description | +|-------|------|-------------| +| **P1** | Left Lobe Transverse | Transverse view of liver left lobe | +| **P2** | Right Lobe Intercostal | Intercostal view of liver right lobe | +| **P3** | Subcostal with Right Kidney | Subcostal view showing liver and right kidney | +| **P4** | Subcostal with Hepatic Vein | Subcostal view showing hepatic vein confluence | + +### Path Segments + +| Segment | Route | Description | +|---------|-------|-------------| +| seg_1 | P1 β†’ P2 | Navigate from left lobe transverse to right lobe intercostal plane | +| seg_2 | P2 β†’ P3 | Navigate from right lobe intercostal to subcostal with right kidney plane | +| seg_3 | P3 β†’ P4 | Navigate from subcostal kidney to subcostal hepatic vein plane | +| seg_4 | P4 β†’ P1 | Navigate from subcostal hepatic vein to left lobe transverse plane | +| seg_5 | P1 β†’ P3 | Navigate diagonal from left lobe to subcostal kidney plane | +| seg_6 | P2 β†’ P4 | Navigate diagonal from right intercostal to hepatic vein plane | +| seg_7 | P2 β†’ P1 | Navigate from right lobe intercostal to left lobe transverse plane | +| seg_8 | P3 β†’ P2 | Navigate from subcostal with right kidney to right lobe intercostal plane | +| seg_9 | P4 β†’ P3 | Navigate from subcostal hepatic vein to subcostal kidney plane | +| seg_10 | P1 β†’ P4 | Navigate from left lobe transverse to subcostal hepatic vein plane | +| seg_11 | P3 β†’ P1 | Navigate diagonal from subcostal kidney to left lobe plane | +| seg_12 | P4 β†’ P2 | Navigate diagonal from hepatic vein to right intercostal plane | + +## State/Action Format + +### observation.state[t] (7D) +| Index | Name | Unit | Description | +|-------|------|------|-------------| +| 0-2 | x, y, z | mm | Probe contact position on skin surface | +| 3-6 | qx, qy, qz, qw | - | Quaternion orientation | + +### action[t] (7D) +Same format as state, representing the **absolute pose at next frame** (state[t+1]). + +## Image Specifications + +| Property | Value | +|----------|-------| +| Format | MP4 video (H.264) | +| Size | 512 Γ— 512 pixels | +| Channels | RGB (3) | +| FPS | 20 Hz | + +## Coordinate System + +- **World coordinates**: Right-handed, units in millimeters (mm) +- **Quaternion**: [qx, qy, qz, qw] format (scipy convention) + +## Synchronization + +All data is frame-synchronized at 20 Hz. Rendering is deterministic. + +## Data Collection + +- Source: Physics-based ultrasound simulation from CT segmentation (3Dircadb dataset) +- Simulation: PyTorch-based acoustic rendering with tissue-specific parameters +- Sampling: 3mm uniform geodesic sampling along skin surface + +## Contact + +For questions or issues, please contact the UltraVision+ Lab. diff --git a/Ultrasound/imfusion/imfusion_dataset/ImFusion_dataset/README.md b/Ultrasound/imfusion/imfusion_dataset/ImFusion_dataset/README.md new file mode 100644 index 0000000000000000000000000000000000000000..aa58932cd0b668d727a781a0dc915478cf2fda27 --- /dev/null +++ b/Ultrasound/imfusion/imfusion_dataset/ImFusion_dataset/README.md @@ -0,0 +1,60 @@ +# IT2S Dataset - README +--- +## Overview +* +The ImFusion Talk2Scan (IT2S) dataset is a multimodal robotic ultrasound dataset designed for learning language-conditioned and skill-conditioned ultrasound scanning behaviors. The dataset combines robot motion data, ultrasound imaging, external visual sensing, and structured task annotations. + +Rather than focusing on long, monolithic procedures, IT2S emphasizes atomic and reusable scanning skills that commonly occur during abdominal ultrasound examinations. This design aims to support research in imitation learning, skill composition, and visual-language-action models for robotic ultrasound.* +--- + +## File Structure +The file structure of this directory is shown below: +``` +ImFusionDataset/ +β”œβ”€β”€ data/ +β”‚ β”œβ”€β”€ chunk-000/ +β”‚ β”‚ β”œβ”€β”€ episode_0000000.parquet +β”‚ β”‚ β”œβ”€β”€ episode_0000001.parquet +β”‚ β”‚ └── ... +β”‚ └── chunk-001/ +β”‚ └── ... +β”œβ”€β”€ videos/ +β”‚ β”œβ”€β”€ chunk-000/ +β”‚ β”‚ β”œβ”€β”€ observation.images.tpv_camera/ +β”‚ β”‚ β”œβ”€β”€ observation.images.ultrasound/ +β”‚ β”‚ └── observation.images.wrist_camera/ +β”‚ └── chunk-001/ +β”‚ └── ... +β”œβ”€β”€ meta/ +β”‚ β”œβ”€β”€ episodes.jsonl +β”‚ β”œβ”€β”€ episodes_stats.jsonl +β”‚ β”œβ”€β”€ tasks.jsonl +β”‚ β”œβ”€β”€ info.json +β”‚ β”œβ”€β”€ stats.json +β”‚ └── README.md +└── +``` +## ⏱️ Data Synchronization Approach +All data streams (robot states and camera feeds) are first converted to a common epoch-based time reference. The synchronization window is determined primarily via motion detection, ensuring that alignment is anchored to the onset of robot motion rather than arbitrary timestamps, with the aim to prevent learning idle states at the begining of the motion or at the end of the motion. If motion detection fails, the system falls back to the maximal overlapping interval across all streams. Within this window, the effective sampling rate of each modality is estimated, and a target frame rate is selected based on the slowest sufficiently reliable (>20Hz) stream. A uniform time grid is then generated. Image modalities are resampled using nearest-neighbor selection to preserve raw frame integrity, while continuous robot signals (e.g., joint states and poses) are linearly interpolated to ensure temporal smoothness. This procedure yields a temporally consistent, multi-modal dataset in which all signals correspond to the same physical time points. + +*Task Design Philosophy* + +The IT2S dataset is intentionally structured around single, well-defined ultrasound skills, rather than full clinical workflows. Each episode typically corresponds to one focused task, enabling precise supervision and compositional learning. + +The primary task categories include: +- Patient approaching +- Losing and re-establishing contact +- Intentional probe detachment and safe re-contact with the surface. +- Horizontal scan +- Lateral to transversal (and vice-verca) probe motion while maintaining surface contact. +- Vertical scan +- Cranio-caudal probe motion with controlled force. +- Fanning +## Attribution & Contact +| | | +| :--- | :--- | +| **Dataset Lead** | `Nikola Budjak` | +| **Institution** | `ImFusion GmbH Munich Germany` | +| **Contact Email** | `budjak@imfusion.com` | +| **Citation (BibTeX)** | 
@misc{robosono_2025,
  author = {Nikola Budjak, Yunye Xiao, Pablo David Aranda Rodriguez, Marco Esposito},
  title = {IT2S: Imfusion's Talk2Scan Language-to-Action Dataset for Robotic Ultrasound Scanning},
  year = {2026},
  publisher = {Open-H-Embodiment},
  note = {}
}
| + diff --git a/Ultrasound/tum/computer_aided_medical_procedures_camp_lab/sonata_all/README.md b/Ultrasound/tum/computer_aided_medical_procedures_camp_lab/sonata_all/README.md new file mode 100644 index 0000000000000000000000000000000000000000..8afc3f4372aba9b076c596c9034328015f21a71f --- /dev/null +++ b/Ultrasound/tum/computer_aided_medical_procedures_camp_lab/sonata_all/README.md @@ -0,0 +1,53 @@ +# SonATA - README + +--- + +## πŸ“‹ At a Glance + +*SonATA is a comprehensive robotic sonography dataset that integrates synchronized ultrasound imaging, external visual data, contact force measurements, robot motion data, and textual instructions, collected from abdomen, thyroid, and arm phantoms to support research on language-conditioned autonomous scanning, perception, and learning-based ultrasound systems.* + +*SonATA, comprising three subset datasets β€” SonATA-Abdomen, SonATA-Thyroid, and SonATA-Arm β€” contains ober 2,000 demonstrations of a robot arm performing diverse ultrasound scanning tasks on three phantoms. It includes three primary task categories (placement, scanning, and navigation), along with **599 unique task instructions** across all subsets, providing a robust dataset for training autonomous robotic ultrasound systems.* + +--- + +## πŸ“Š Dataset Statistics + +| Subset | Episodes | Frames | Tasks | Duration | Train | Val | Test | +| :--- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | +| **SonATA-Arm** | 605 | 265,451 | 195 | 147.5 min (2.46 h) | 423 | 60 | 122 | +| **SonATA-Abdomen** | 1,532 | 325,594 | 284 | 180.9 min (3.01 h) | 1,070 | 154 | 308 | +| **SonATA-Thyroid** | 260 | 42,559 | 120 | 23.6 min (0.39 h) | 182 | 24 | 54 | +| **SonATA-All** | **2,397** | **633,604** | **599** | **352.0 min (5.87 h)** | **1,677** | **238** | **482** | + +- **FPS**: 30 +- **Train / Val / Test split**: 70% / 10% / 20% (per-subset stratified shuffle) + +--- + +## πŸ“– File Structure +The file structure of this directory is shown below: +```text +./ +β”œβ”€β”€ SonATA_*/ +β”‚ └── ... +β”œβ”€β”€ overview/ +β”‚ β”œβ”€β”€ SonATA_*/ +β”‚ β”œβ”€β”€ SonATA_*.csv +β”‚ └── lerobot_config.yaml +└── README.md +``` + +The `SonATA_*` directories contain all demonstrated data, with each directory corresponding to a specific SonATA subset. The `overview/` directory provides the initial configurations for each demonstration in the form of images, together with the corresponding textual instructions and trajectory lengths. The features of the converted LeRobot dataset are defined in `lerobot_config.yaml`. + +## πŸ‘₯ Attribution & Contact + +*Please provide attribution for the dataset creators and a point of contact.* + +| | | +| :--- | :--- | +| **Dataset Lead** | `Dianye Huang` | +| **Institution** | `Technical University of Munich (TUM), Germany` | +| **Contact Email** | `dianye.huang@tum.de` | +| **Citation (BibTeX)** | 
@misc{robosono_2025,
  author = {Dianye Huang, Pei Liu, Zhongliang Jiang, Nassir Navab},
  title = {SonATA: A Robotic Sonography Dataset for Abdomen, Thyroid and Arm Examination},
  year = {2025},
  publisher = {Open-H-Embodiment},
  note = {}
}
| + +