Fix urdf extraction from shapenet dataset, add instructions to README

Author: maurock

.gitignore

@@ -3,7 +3,7 @@
 **/google_scanned_objects/
 **/ycb_objects/
 **/test_folder/
-**/shapenet/
+**/shapenet/ShapeNetCore.v2
 __pycache__
 *.egg-info
 *.bak

README.md

@@ -24,6 +24,7 @@ The list of available objects includes:
 - [Google Scanned Objects](#google-scanned-objects)
 - [Evolved Grasping Analysis Dataset (EGAD)](#egad-objects)
 - [Gibson Objects](#gibson-objects)
+- [ShapeNetCore.v2](#shapenetcorev2)
 
 Each object has the following files:
 
@@ -178,6 +179,36 @@ To set up these objects:
 
  - Some of the meshes have been simplified to reduce the number of triangles.
 
+### ShapeNetCore.v2 ###
+
+- The dataset is available on the ShapeNet website ([https://shapenet.org/](https://shapenet.org/)). 
+
+To extract `.urdf` files from the `ShapeNetCore.v2` dataset, please follow these instructions:
+
+1. Download the `ShapeNetCore.v2` dataset and place it under `pybullet_object_models/shapenet/`. The name of the dataset folder needs to be `ShapeNetCore.v2`. 
+2. Copy the `_protype.urdf` file from `/egad/` into the `ShapeNetCore.v2` directory.
+3. Once these two steps are done, you should have the following tree:
+```
+pybullet_object_models
+ ├── shapenet
+ │   ├── ShapeNetCore.v2
+ │   │   ├── _prototype.urdf
+ │   │   ├── ...   # example: 02942699 
+ |   |   |   ├── ...    # example: 1ab3abb5c090d9b68e940c4e64a94e1e
+```
+4. In the project root folder, run `python build_shapenet_urdfs.py`. The extracted `.urdf` files are named `model.urdf`. They are placed under the object ID directory. For example:
+```
+pybullet_object_models
+ ├── shapenet
+ │   ├── ShapeNetCore.v2
+ │   │   ├── _prototype.urdf
+ │   │   ├── 02942699 
+ |   |   |   ├── 1ab3abb5c090d9b68e940c4e64a94e1e
+ |   |   |   |   ├── models
+ |   |   |   |   ├── images
+ |   |   |   |   ├── model.urdf <-- EXTRACTED .URDF 
+```
+
 ### Usage ###
 
 Example scripts for importing the objects into pybullet are provided in the examples folder. To run these cd into the `examples` directory and run `python demo_load_object.py`. Use the `-object_set=` argument to load from a given object set, currently this can be selected from `primitive`, `random`, `ycb`, `superquadric` or `google` if setup correctly.

pybullet_object_models/build_shapenet_urdfs.py

@@ -7,7 +7,7 @@
 # Build entire libraries of URDFs
 # Can take a while
 
-shapenet_folder = "shapenet/ShapeNetCoreV2"
+shapenet_folder = "shapenet/ShapeNetCore.v2"
 
 subdirs = get_immediate_subdirectories(shapenet_folder)
 
@@ -24,7 +24,7 @@
     builder.build_library(force_overwrite=True,
                           decompose_concave=False,
                           force_decompose=False,
-                          center='mass')
+                          center='geometric')
 
     # tidy up
     move_urdfs_to_subdirs(object_folder)