Awesome
This is a modification of GG-CNN
Note: This repository is based on the original implementation. The original implementation is here.
Modifications
- Add GraspNet 1Billion dataset. The dataset can be downloaded here.
- Add deeper GGCNN network.
Usage
-
You should download the GraspNet 1Billion Dataset first. (abbre. gn1b)
-
Install graspnetAPI following here.
pip install graspnetAPI
-
Change the
graspnet_root
in scriptconvert_gn1b_2_cornell.py
, then run it to convert the dataset format into Cornell Dataset format. -
Run the training script
train_ggcnn.py
to train the model using gn1b Dataset.Example
python train_ggcnn.py --network ggcnn2 --dataset graspnet1b --dataset-path path-to-gn1b --camera realsense --epochs 20 --batch-size 16 --num-workers 20 --description ggcnn2_gn1b