Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -1,48 +1,48 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: cc-by-4.0
|
| 3 |
-
task_categories:
|
| 4 |
-
- robotics
|
| 5 |
-
tags:
|
| 6 |
-
-
|
| 7 |
-
- manipulation
|
| 8 |
-
- lerobot
|
| 9 |
-
- franka-panda
|
| 10 |
-
- iphone-lidar
|
| 11 |
-
size_categories:
|
| 12 |
-
- n<1K
|
| 13 |
-
---
|
| 14 |
-
|
| 15 |
-
#
|
| 16 |
-
|
| 17 |
-
Robot training dataset generated from a single iPhone 15 Pro LiDAR recording of a human picking up a mug.
|
| 18 |
-
|
| 19 |
-
## Pipeline
|
| 20 |
-
iPhone .r3d recording -> GroundingDINO (object detection) -> HaMeR (3D hand mesh) -> LiDAR depth fusion -> Spatial trajectory retargeting (V5 hybrid) -> IK solve -> LeRobot HDF5
|
| 21 |
-
|
| 22 |
-
## Results
|
| 23 |
-
- **18.1cm clean mug lift** in MuJoCo Franka Panda simulation
|
| 24 |
-
- 0 table collisions
|
| 25 |
-
- 93.7% hand pose coverage (282/301 frames)
|
| 26 |
-
|
| 27 |
-
## Data Format
|
| 28 |
-
- **Observations** (14-dim): qpos_arm[7] + ee_pos[3] + mug_pos[3] + gripper[1]
|
| 29 |
-
- **Actions** (8-dim): target_qpos[7] + target_gripper[1]
|
| 30 |
-
- **Episodes**: 1 (270 steps @ 30fps)
|
| 31 |
-
|
| 32 |
-
## Usage
|
| 33 |
-
```python
|
| 34 |
-
import h5py
|
| 35 |
-
with h5py.File("data/data.hdf5", "r") as f:
|
| 36 |
-
ep = f["episode_0"]
|
| 37 |
-
obs = ep["observations/qpos_arm"][:]
|
| 38 |
-
actions = ep["actions/target_qpos"][:]
|
| 39 |
-
```
|
| 40 |
-
|
| 41 |
-
## Source
|
| 42 |
-
- Recording: iPhone 15 Pro, Record3D app
|
| 43 |
-
- Robot: Franka Emika Panda 7-DoF
|
| 44 |
-
- Simulator: MuJoCo
|
| 45 |
-
- Pipeline: [
|
| 46 |
-
|
| 47 |
-
## License
|
| 48 |
-
CC-BY-4.0
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-4.0
|
| 3 |
+
task_categories:
|
| 4 |
+
- robotics
|
| 5 |
+
tags:
|
| 6 |
+
- FlexaTechnologies
|
| 7 |
+
- manipulation
|
| 8 |
+
- lerobot
|
| 9 |
+
- franka-panda
|
| 10 |
+
- iphone-lidar
|
| 11 |
+
size_categories:
|
| 12 |
+
- n<1K
|
| 13 |
+
---
|
| 14 |
+
|
| 15 |
+
# Flexa Pick Mug V5
|
| 16 |
+
|
| 17 |
+
Robot training dataset generated from a single iPhone 15 Pro LiDAR recording of a human picking up a mug.
|
| 18 |
+
|
| 19 |
+
## Pipeline
|
| 20 |
+
iPhone .r3d recording -> GroundingDINO (object detection) -> HaMeR (3D hand mesh) -> LiDAR depth fusion -> Spatial trajectory retargeting (V5 hybrid) -> IK solve -> LeRobot HDF5
|
| 21 |
+
|
| 22 |
+
## Results
|
| 23 |
+
- **18.1cm clean mug lift** in MuJoCo Franka Panda simulation
|
| 24 |
+
- 0 table collisions
|
| 25 |
+
- 93.7% hand pose coverage (282/301 frames)
|
| 26 |
+
|
| 27 |
+
## Data Format
|
| 28 |
+
- **Observations** (14-dim): qpos_arm[7] + ee_pos[3] + mug_pos[3] + gripper[1]
|
| 29 |
+
- **Actions** (8-dim): target_qpos[7] + target_gripper[1]
|
| 30 |
+
- **Episodes**: 1 (270 steps @ 30fps)
|
| 31 |
+
|
| 32 |
+
## Usage
|
| 33 |
+
```python
|
| 34 |
+
import h5py
|
| 35 |
+
with h5py.File("data/data.hdf5", "r") as f:
|
| 36 |
+
ep = f["episode_0"]
|
| 37 |
+
obs = ep["observations/qpos_arm"][:]
|
| 38 |
+
actions = ep["actions/target_qpos"][:]
|
| 39 |
+
```
|
| 40 |
+
|
| 41 |
+
## Source
|
| 42 |
+
- Recording: iPhone 15 Pro, Record3D app
|
| 43 |
+
- Robot: Franka Emika Panda 7-DoF
|
| 44 |
+
- Simulator: MuJoCo
|
| 45 |
+
- Pipeline: [FlexaTechnologies-pipeline](https://github.com/ChristianNyamekye/FlexaTechnologies-pipeline)
|
| 46 |
+
|
| 47 |
+
## License
|
| 48 |
+
CC-BY-4.0
|