Modified code for GraspNet Baseline

Baseline model for "GraspNet-1Billion: A Large-Scale Benchmark for General Object Grasping" (CVPR 2020).

Requirements

• Python 3
• PyTorch 1.6
• Open3d >=0.8
• TensorBoard 2.3
• NumPy
• SciPy
• Pillow
• tqdm

Installation

Get the code.

git clone https://github.com/graspnet/graspnet-baseline.git
cd graspnet-baseline

Install packages via Pip.

pip install -r requirements.txt

Compile and install pointnet2 operators (code adapted from votenet).

cd pointnet2
python setup.py install

Compile and install knn operator (code adapted from pytorch_knn_cuda).

cd knn
python setup.py install

Install graspnetAPI for evaluation.

git clone https://github.com/graspnet/graspnetAPI.git
cd graspnetAPI
pip install .

When using, please use these checkpoints

• checkpoint-rs.tar [Google Drive] [Baidu Pan]
• checkpoint-kn.tar [Google Drive] [Baidu Pan]

checkpoint-rs.tar and checkpoint-kn.tar are trained using RealSense data and Kinect data respectively.

Main changes

Camera

A class for camera control was created as RealSenseCameraV2.py

Graspnet

A modified Demo.py was created and named GraspPointsFromRealsense.py. This modified demo.py is a class that outputs grasping poses with an associated grasp score using the RealSenseCameraV2 as input.

Testing

The file Main_using_urx.py contains the prototype for grasping. It uses the urx package to control the ur robot