Dataset card: meshes and annotations

README.md

@@ -9,22 +9,21 @@ tags:
   - language-conditioned
   - 3d
   - meshes
-  - point-cloud
   - digital-twin
 size_categories:
-  - 1K<n<10K
+  - n<1K
 ---
 
 # GraspTwin benchmark scenes
 
-Real tabletop scenes captured for language-conditioned grasping, each paired with a
-digital twin: one reconstructed `.glb` mesh per object plus the object poses needed to
-rebuild the scene in simulation.
+Digital twins of real tabletop scenes for language-conditioned grasping: one
+reconstructed `.glb` mesh per object, the object poses needed to rebuild each scene in
+simulation, and the natural-language tasks with their target objects.
 
 - 9 scenes, 5 subscenes each (45 subscenes total)
 - 8 household objects per subscene (`scene_8_4` has 10)
 - 10 natural-language tasks per scene, each tied to a target object
-- about 5.6 GB in total (meshes 2.6 GB, multi-view captures 2.2 GB, point clouds 0.8 GB)
+- about 2.6 GB of meshes
 
 ## Layout
 
@@ -34,19 +33,12 @@ scene_N/
   scene_N_M/                     # subscene M = 1..5 (same objects, different arrangement)
     objects.json                 # names of the objects in the subscene
     target_objects.json          # target object per task, index-aligned with ../tasks.json
-    boxes.json                   # 2D box [x1, y1, x2, y2] in rgb.png pixels, per object name
-    rgb.png, depth.png           # 1920x1080 scene capture
-    pointcloud.ply               # scene point cloud (xyz, ~1.5M points)
+    boxes.json                   # 2D box [x1, y1, x2, y2] in pixels of the 1920x1080
+                                 #   scene capture, per object name
     meshes/
       <object name>.glb          # one mesh per object (trimesh export)
       transforms.json            # per-object translation, rotation_wxyz, scale
       settle_transforms.json     # same, after the objects settle under physics in sim
-    obs/
-      obs/                       # single-view capture: rgb.png, pc.npy, transforms.json
-                                 #   (camera intrinsics and camera_to_robot extrinsics)
-      lerf_obs/                  # 60-frame multi-view capture: images/frame_*.jpg and
-                                 #   a nerfstudio-style transforms.json
-      validation_obs/            # left.png, right.png (1280x720)
 ```
 
 Object names are the keys used across `objects.json`, `boxes.json`, `transforms.json`