Update README.md
Browse files
README.md
CHANGED
|
@@ -16,7 +16,7 @@ ManipGen-UniDexGrasp is used to train pick and place policies in [ManipGen](http
|
|
| 16 |
3. Sample grasp pose: We sample grasp poses for each object using antipodal sampling. Then we teleport the gripper to grasp the object from the sampled poses, add perturbations, and record valid grasp poses. The ratio of valid grasp pose is also used to filter intractable rest poses (e.g., an upside-down bowl).
|
| 17 |
|
| 18 |
### Structure
|
| 19 |
-
* `datasetv4.1_posedata.npy`: rest poses
|
| 20 |
* `meshdatav3_scaled/`: mesh and urdf
|
| 21 |
-
* `graspdata/`: pre-sampled grasp poses
|
| 22 |
* `trainset3419.txt`: the list of 3419 objects used to train pick and place policies in ManipGen in the format of "[object_code] [object_scale]"
|
|
|
|
| 16 |
3. Sample grasp pose: We sample grasp poses for each object using antipodal sampling. Then we teleport the gripper to grasp the object from the sampled poses, add perturbations, and record valid grasp poses. The ratio of valid grasp pose is also used to filter intractable rest poses (e.g., an upside-down bowl).
|
| 17 |
|
| 18 |
### Structure
|
| 19 |
+
* `datasetv4.1_posedata.npy`: rest poses for each object
|
| 20 |
* `meshdatav3_scaled/`: mesh and urdf
|
| 21 |
+
* `graspdata/`: pre-sampled grasp poses for Franka arm with UMI gripper
|
| 22 |
* `trainset3419.txt`: the list of 3419 objects used to train pick and place policies in ManipGen in the format of "[object_code] [object_scale]"
|