WIP object and grasping dataset
Code for PAPER
Requirements can be installed from requirements.txt. This can be done by running the following commands:
python3 -m pip install --upgrade pip
pip3 install -r requirements.txt
Data can be either directly downloaded from Monash Bridges or using the provided python script.
Python: Once the repo has been cloned, from the root directory run:
python scripts/download_object_dataset.py --object --grasp --process_grasp
This will download both the object and grasping datasets, and preprocess the grasping dataset. If you wish to not download one of the datasets or not to preprocess the dataset, please delete the corresponding arguments:
- download object dataset argument: --object
- download grasp dataset argument: --grasp
- preprocess grasp dataset argument: --process_grasp
Monash Bridges: Datasets can be downloaded from the following links (Can easily be downloaded by selecting 'Download all' on the webpages):
To preprocess the grasping dataset, ensure the grasping dataset is modified to have the following file structure:
project
|
|--grasp_ds
| |--json_files
| | |--[1-20]
| | | |--[0-74]
| |--depth_images
| | |--[1-20]
| | | |--[0-74].png
| |--rgb_images
| | |--[1-20]
| | | |--[0-74].png
| |--point_clouds
| | |--[1-20]
| | | |--[0-74].pcd
where the folders named 1 to 20 in point_clouds_00X have been moved into point_clouds. Then run:
python scripts/process_clouds.py
- To load some sample data, run
python scripts/data_loading_tools.py
- To visualise the point cloud through the data processing process, run
python scripts/visualise_data_processing.py
- To visualise the grasp pose, run
python scripts/display_grasp_pose.py
- To preform registration between a point cloud and it's corresponding object mesh, run
python scripts/grasp_pcl_object_registration.py
WIP