Update README.md
Browse files
README.md
CHANGED
|
@@ -72,13 +72,6 @@ This release contains **only dozens of sample trajectories** and is intended for
|
|
| 72 |
- Preliminary algorithm development
|
| 73 |
- Demonstrating multimodal alignment and synchronization
|
| 74 |
|
| 75 |
-
Included sensor streams:
|
| 76 |
-
|
| 77 |
-
- RGB camera images
|
| 78 |
-
- Merged_Trajectory
|
| 79 |
-
- ToF point cloud frames
|
| 80 |
-
- Clamp (gripper width) sensor readings
|
| 81 |
-
- Fused multi-sensor trajectory
|
| 82 |
|
| 83 |
Full-scale datasets are available upon request for research or enterprise collaboration.
|
| 84 |
|
|
@@ -87,55 +80,16 @@ Full-scale datasets are available upon request for research or enterprise collab
|
|
| 87 |
|
| 88 |
## 📊 Data Specifications
|
| 89 |
|
|
|
|
|
|
|
| 90 |
| **Data Type** | **Path** | **Shape** | **Type** | **Description** |
|
| 91 |
|--------------|----------|-----------|----------|-----------------|
|
| 92 |
-
| RGB Images | RGB_Images/Frames/*.
|
| 93 |
| ToF PointClouds | ToF_PointClouds/PointClouds/*.pcd | variable | pcd | Time-of-Flight point clouds |
|
| 94 |
| Clamp Data | Clamp_Data/clamp_data_tum.txt | (N, 2) | float | Timestamp + clamp width |
|
| 95 |
| Merged Trajectory | Merged_Trajectory/merged_trajectory.txt | (N, 8) | float | Fused multi-sensor pose |
|
| 96 |
|
| 97 |
-
---
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
## 🧭 Pose Data Format
|
| 101 |
-
|
| 102 |
-
All pose data (SLAM, Vive, fused) follow the same structure:
|
| 103 |
-
|
| 104 |
-
```markdown
|
| 105 |
-
timestamp x y z qx qy qz qw
|
| 106 |
-
|
| 107 |
-
```
|
| 108 |
-
|
| 109 |
-
| Field | Description | Field | Description |
|
| 110 |
-
| :---: | :---: | :---: | :---: |
|
| 111 |
-
| timestamp | Unix timestamp | qx | Quaternion X component |
|
| 112 |
-
| x | Position X (meters) | qy | Quaternion Y component |
|
| 113 |
-
| y | Position Y (meters) | qz | Quaternion Z component |
|
| 114 |
-
| z | Position Z (meters) | qw | Quaternion W component |
|
| 115 |
-
|
| 116 |
-
### Coordinate System Reference
|
| 117 |
-
|
| 118 |
-
To ensure correct visualization and control, please adhere to the following frame definition:
|
| 119 |
-
|
| 120 |
-
* **Origin (0,0,0):** Geometric center of the tracking base stations (World Frame).
|
| 121 |
-
* 🔴 **X-Axis:** Points **Right** (facing the robot workspace).
|
| 122 |
-
* 🟢 **Y-Axis:** Points **Forward** (main manipulation direction).
|
| 123 |
-
* 🔵 **Z-Axis:** Points **Upward** (aligned with gravity, opposing g).
|
| 124 |
-
|
| 125 |
-
### Data Structure
|
| 126 |
-
|
| 127 |
-
* **Timestamp:** Unix timestamp (seconds).
|
| 128 |
-
* **Position (x, y, z):** Translation vector in meters.
|
| 129 |
-
* **Orientation (qx, qy, qz, qw):** Rotation quaternion (Scalar-last format).
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
<div align="center">
|
| 133 |
-
<small><i>Tip: If using ROS or specific simulators (like Isaac Gym), ensure you apply the necessary transform if your Z-axis convention differs.</i></small>
|
| 134 |
-
</div>
|
| 135 |
-
|
| 136 |
-
---
|
| 137 |
|
| 138 |
-
## 🗂️ Dataset Structure
|
| 139 |
|
| 140 |
**Purpose:** The dataset is organized by tasks and sessions, with each task containing multimodal sensor data streams captured during robotic manipulation episodes. Below is the hierarchical structure observed from the actual dataset:
|
| 141 |
|
|
@@ -143,7 +97,7 @@ To ensure correct visualization and control, please adhere to the following fram
|
|
| 143 |
📂 dataset_root/
|
| 144 |
├── 🥤 pick_coke/ # [Task] Picking up a coke can
|
| 145 |
│ ├── 📷 RGB_Images/
|
| 146 |
-
│ │ └── 🎞️ Frames/ # Multi-view RGB image frames (.
|
| 147 |
│ ├── ☁️ ToF_PointClouds/
|
| 148 |
│ │ └── 🧊 PointClouds/ # Depth sensor point cloud data (.pcd)
|
| 149 |
│ ├── 🔧 Clamp_Data/ # Gripper width measurements (.txt)
|
|
@@ -159,48 +113,48 @@ To ensure correct visualization and control, please adhere to the following fram
|
|
| 159 |
└── 📁 session_folders/ # Individual recording sessions
|
| 160 |
```
|
| 161 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 162 |
---
|
| 163 |
|
| 164 |
-
|
| 165 |
-
|
| 166 |
-
|
| 167 |
-
|
| 168 |
-
|
| 169 |
-
* **
|
| 170 |
-
* **
|
| 171 |
-
|
| 172 |
-
|
| 173 |
-
|
| 174 |
-
|
| 175 |
-
|
| 176 |
-
|
| 177 |
-
|
| 178 |
-
|
| 179 |
-
|
| 180 |
-
|
| 181 |
-
3. `Merged_Trajectory/`
|
| 182 |
-
* **Description:** Fused trajectory data combining SLAM, Vive tracking, and IMU inputs. Represents the 6-DoF pose of the end-effector.
|
| 183 |
-
* **File:** `merged_trajectory.txt`
|
| 184 |
-
* **Shape:** `(num_timesteps, 8)`
|
| 185 |
-
* **Columns:**
|
| 186 |
-
*
|
| 187 |
-
```text
|
| 188 |
-
[Timestamp, Pos_X, Pos_Y, Pos_Z, Q_X, Q_Y, Q_Z, Q_W]
|
| 189 |
-
```
|
| 190 |
-
|
| 191 |
-
* `Timestamp`: Unix timestamp (seconds)
|
| 192 |
-
* `Pos`: Cartesian position (meters)
|
| 193 |
-
* `Q`: Rotation quaternion (scalar-last `xyzw`)
|
| 194 |
-
|
| 195 |
-
4. `Clamp_Data/`
|
| 196 |
-
* **Description:** Records the state of the end-effector (gripper) over time.
|
| 197 |
-
* **File:** `clamp_data_tum.txt`
|
| 198 |
-
* **Shape:** `(num_timesteps, 2)`
|
| 199 |
-
* **Columns:**
|
| 200 |
-
```text
|
| 201 |
-
[Timestamp, Width]
|
| 202 |
-
```
|
| 203 |
-
* `Width`: Opening distance of the clamp (meters/units).
|
| 204 |
|
| 205 |
---
|
| 206 |
|
|
|
|
| 72 |
- Preliminary algorithm development
|
| 73 |
- Demonstrating multimodal alignment and synchronization
|
| 74 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 75 |
|
| 76 |
Full-scale datasets are available upon request for research or enterprise collaboration.
|
| 77 |
|
|
|
|
| 80 |
|
| 81 |
## 📊 Data Specifications
|
| 82 |
|
| 83 |
+
|
| 84 |
+
|
| 85 |
| **Data Type** | **Path** | **Shape** | **Type** | **Description** |
|
| 86 |
|--------------|----------|-----------|----------|-----------------|
|
| 87 |
+
| RGB Images | RGB_Images/Frames/*.mp4| (H, W, 3) | uint8 | Multi-view RGB images |
|
| 88 |
| ToF PointClouds | ToF_PointClouds/PointClouds/*.pcd | variable | pcd | Time-of-Flight point clouds |
|
| 89 |
| Clamp Data | Clamp_Data/clamp_data_tum.txt | (N, 2) | float | Timestamp + clamp width |
|
| 90 |
| Merged Trajectory | Merged_Trajectory/merged_trajectory.txt | (N, 8) | float | Fused multi-sensor pose |
|
| 91 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 92 |
|
|
|
|
| 93 |
|
| 94 |
**Purpose:** The dataset is organized by tasks and sessions, with each task containing multimodal sensor data streams captured during robotic manipulation episodes. Below is the hierarchical structure observed from the actual dataset:
|
| 95 |
|
|
|
|
| 97 |
📂 dataset_root/
|
| 98 |
├── 🥤 pick_coke/ # [Task] Picking up a coke can
|
| 99 |
│ ├── 📷 RGB_Images/
|
| 100 |
+
│ │ └── 🎞️ Frames/ # Multi-view RGB image frames (.mp4)
|
| 101 |
│ ├── ☁️ ToF_PointClouds/
|
| 102 |
│ │ └── 🧊 PointClouds/ # Depth sensor point cloud data (.pcd)
|
| 103 |
│ ├── 🔧 Clamp_Data/ # Gripper width measurements (.txt)
|
|
|
|
| 113 |
└── 📁 session_folders/ # Individual recording sessions
|
| 114 |
```
|
| 115 |
|
| 116 |
+
* Sensor Data Details:
|
| 117 |
+
* RGB_Images/video.mp4: A video file containing multi-view RGB footage of the manipulation task.
|
| 118 |
+
* ToF_PointClouds/: Contains dense 3D point cloud data (.pcd files) from Time-of-Flight sensors.
|
| 119 |
+
* Merged_Trajectory/merged_trajectory.txt: The primary 6-DoF end-effector trajectory file, fusing multiple sensor inputs.
|
| 120 |
+
* Clamp_Data/clamp_data_tum.txt: Records the gripper's opening width over time.
|
| 121 |
+
---
|
| 122 |
+
|
| 123 |
+
|
| 124 |
+
## 🧭 Data Formats
|
| 125 |
+
|
| 126 |
+
All pose data (SLAM, Vive, fused) follow the same structure:
|
| 127 |
+
|
| 128 |
+
```markdown
|
| 129 |
+
timestamp x y z qx qy qz qw
|
| 130 |
+
|
| 131 |
+
```
|
| 132 |
+
|
| 133 |
+
| Field | Description | Field | Description |
|
| 134 |
+
| :---: | :---: | :---: | :---: |
|
| 135 |
+
| timestamp | Unix timestamp | qx | Quaternion X component |
|
| 136 |
+
| x | Position X (meters) | qy | Quaternion Y component |
|
| 137 |
+
| y | Position Y (meters) | qz | Quaternion Z component |
|
| 138 |
+
| z | Position Z (meters) | qw | Quaternion W component |
|
| 139 |
+
|
| 140 |
---
|
| 141 |
|
| 142 |
+
### Coordinate System
|
| 143 |
+
|
| 144 |
+
To ensure correct visualization and control, all pose data adheres to the following right-handed coordinate system (World Frame).
|
| 145 |
+
* **Origin (0,0,0):** Geometric center of the tracking base stations (World Frame).
|
| 146 |
+
* 🔴 **X-Axis:** Points Right (relative to the workspace).
|
| 147 |
+
* 🟢 **Y-Axis:** Points Forward (the primary direction of manipulation).
|
| 148 |
+
* 🔵 **Z-Axis:** Points Upward (opposite to the direction of gravity).
|
| 149 |
+
<div align="center">
|
| 150 |
+
<img src="https://huggingface.co/datasets/FastUMIPro/example_data_fastumi_pro_raw/resolve/main/1_2025-12-11_201019_400.jpg" alt="Coordinate System Visualization" width="50%" style="border-radius: 8px; margin-top: 15px;">
|
| 151 |
+
<br>
|
| 152 |
+
<small><i>Visual reference for the coordinate system.</i></small>
|
| 153 |
+
</div>
|
| 154 |
+
<div align="center" style="margin-top: 15px;">
|
| 155 |
+
<small><i>Tip: When using simulation environments like ROS or Isaac Gym, ensure your coordinate frame conventions match. You may need to apply a transformation if your framework uses a different "up" axis (e.g., Z-up vs. Y-up).</i></small>
|
| 156 |
+
</div>
|
| 157 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 158 |
|
| 159 |
---
|
| 160 |
|