Hui commited on
Commit
63f3695
·
1 Parent(s): 6606a71

Add MANO tabletop archive

Browse files
Files changed (2) hide show
  1. README.md +25 -6
  2. mano_tabletop.zip +3 -0
README.md CHANGED
@@ -69,6 +69,7 @@ The repository contains the following archive groups:
69
  | `allegro_dataset_1.zip` | Allegro Hand grasping motion sequences. |
70
  | `allegro_dataset_2.zip` | Additional Allegro Hand grasping motion sequences. |
71
  | `allegro_tabletop.zip` | Allegro Hand tabletop grasping motions. Objects are placed on a table and the hand approaches from above. |
 
72
  | `sharpa_tabletop.zip` | SharpA Hand tabletop grasping motions. Objects are placed on a table and the hand approaches from above. |
73
  | `mano_dataset_1.zip.part00`, `mano_dataset_1.zip.part01` | Split parts of `mano_dataset_1.zip`, containing MANO hand grasping motion sequences. |
74
  | `mano_dataset_2.zip.part00`, `mano_dataset_2.zip.part01` | Split parts of `mano_dataset_2.zip`, containing additional MANO hand grasping motion sequences. |
@@ -152,12 +153,12 @@ recorded motion sequences.
152
 
153
  ## Tabletop sequences
154
 
155
- `allegro_tabletop.zip` and `sharpa_tabletop.zip` contain grasping motions in a
156
- tabletop setting. In the original diverse-motion archives, the hand approaches
157
- objects from different directions to generate broad hand-object interaction
158
- coverage. In the tabletop archives, each object is placed on a table and the hand
159
- approaches the object from above. This setting is useful for applications that
160
- focus on tabletop grasping scenes.
161
 
162
  The tabletop archives are organized as follows:
163
 
@@ -201,10 +202,14 @@ sharpa_tabletop.zip
201
  `-- ...
202
  ```
203
 
 
 
 
204
  Each tabletop `.npy` file contains a single motion sequence:
205
 
206
  ```python
207
  data = np.load("allegro_0.npy", allow_pickle=True).item()
 
208
  data = np.load("sharpa_0.npy", allow_pickle=True).item()
209
  ```
210
 
@@ -221,6 +226,20 @@ For Allegro tabletop sequences:
221
  - `data[object_id]["rot"]`: NumPy array of shape `(frame_num, 3)`. Object
222
  orientation sequence in axis-angle representation.
223
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
224
  For SharpA tabletop sequences:
225
 
226
  - `data["right_hand"]["trans"]`: NumPy array of shape `(frame_num, 3)`. Wrist
 
69
  | `allegro_dataset_1.zip` | Allegro Hand grasping motion sequences. |
70
  | `allegro_dataset_2.zip` | Additional Allegro Hand grasping motion sequences. |
71
  | `allegro_tabletop.zip` | Allegro Hand tabletop grasping motions. Objects are placed on a table and the hand approaches from above. |
72
+ | `mano_tabletop.zip` | MANO hand tabletop grasping motions. Objects are placed on a table and the hand approaches from above. |
73
  | `sharpa_tabletop.zip` | SharpA Hand tabletop grasping motions. Objects are placed on a table and the hand approaches from above. |
74
  | `mano_dataset_1.zip.part00`, `mano_dataset_1.zip.part01` | Split parts of `mano_dataset_1.zip`, containing MANO hand grasping motion sequences. |
75
  | `mano_dataset_2.zip.part00`, `mano_dataset_2.zip.part01` | Split parts of `mano_dataset_2.zip`, containing additional MANO hand grasping motion sequences. |
 
153
 
154
  ## Tabletop sequences
155
 
156
+ `allegro_tabletop.zip`, `mano_tabletop.zip`, and `sharpa_tabletop.zip` contain
157
+ grasping motions in a tabletop setting. In the original diverse-motion archives,
158
+ the hand approaches objects from different directions to generate broad
159
+ hand-object interaction coverage. In the tabletop archives, each object is placed
160
+ on a table and the hand approaches the object from above. This setting is useful
161
+ for applications that focus on tabletop grasping scenes.
162
 
163
  The tabletop archives are organized as follows:
164
 
 
202
  `-- ...
203
  ```
204
 
205
+ `mano_tabletop.zip` follows the same directory structure and contains
206
+ `mano_0.npy`, `mano_1.npy`, ... sequence files.
207
+
208
  Each tabletop `.npy` file contains a single motion sequence:
209
 
210
  ```python
211
  data = np.load("allegro_0.npy", allow_pickle=True).item()
212
+ data = np.load("mano_0.npy", allow_pickle=True).item()
213
  data = np.load("sharpa_0.npy", allow_pickle=True).item()
214
  ```
215
 
 
226
  - `data[object_id]["rot"]`: NumPy array of shape `(frame_num, 3)`. Object
227
  orientation sequence in axis-angle representation.
228
 
229
+ For MANO tabletop sequences:
230
+
231
+ - `data["right_hand"]["trans"]`: NumPy array of shape `(frame_num, 3)`. Wrist
232
+ position sequence.
233
+ - `data["right_hand"]["rot"]`: NumPy array of shape `(frame_num, 3)`. Wrist
234
+ orientation sequence in axis-angle representation. For MANO, these are the
235
+ first 3 dimensions of the MANO parameter.
236
+ - `data["right_hand"]["pose"]`: NumPy array of shape `(frame_num, 45)`. The
237
+ remaining 45 dimensions of the MANO parameter.
238
+ - `data[object_id]["trans"]`: NumPy array of shape `(frame_num, 3)`. Object
239
+ position sequence.
240
+ - `data[object_id]["rot"]`: NumPy array of shape `(frame_num, 3)`. Object
241
+ orientation sequence in axis-angle representation.
242
+
243
  For SharpA tabletop sequences:
244
 
245
  - `data["right_hand"]["trans"]`: NumPy array of shape `(frame_num, 3)`. Wrist
mano_tabletop.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4ce4f884892147b1419716def4813010f011c5b8bf67a1f15db7fcee4c55142c
3
+ size 18920168549