squeeze-ggcnn-master

Note: The program was modified by reference to Closing the Loop for Robotic Grasping.

This repository contains the implementation of the Generative Grasping Convolutional Neural Network (GG-CNN) from the paper:

Closing the Loop for Robotic Grasping: A Real-time, Generative Grasp Synthesis Approach

Douglas Morrison, Peter Corke, Jürgen Leitner

Robotics: Science and Systems (RSS) 2018

arXiv | Video

@inproceedings{morrison2018closing, title={{Closing the Loop for Robotic Grasping: A Real-time, Generative Grasp Synthesis Approach}}, author={Morrison, Douglas and Corke, Peter and Leitner, J"urgen}, booktitle={Proc.\ of Robotics: Science and Systems (RSS)}, year={2018} }

Generative Grasping CNN (squeeze)

The squeeze is a lightweight, fully-convolutional network which predicts the quality and pose of antipodal grasps at every pixel in an input depth image. The lightweight and single-pass generative nature of squeeze allows for fast execution and closed-loop control, enabling accurate grasping in dynamic environments where objects are moved during the grasp attempt.

Installation

Operating system：Ubuntu MATE16.04 CPU：Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10GHz GPU：Titan X python版本：3.7.13 torch版本：1.10.1+cu111 torchvision版本：0.11.2++cu111

This code was developed with Python 3.7 on Ubuntu 16.04. Python requirements can installed by:

pip install -r requirements.txt

Datasets

Currently, both the Cornell Grasping Dataset and Jacquard Dataset are supported.

Cornell Grasping Dataset

1. Download the and extract Cornell Grasping Dataset.
2. Convert the PCD files to depth images by running python -m utils.dataset_processing.generate_cornell_depth <Path To Dataset>

（官方链接已失效，此链接为csdn其他人保存的数据集副本） cornell数据集：https://blog.csdn.net/qq_40081208/article/details/104547339

Jacquard Dataset

1. Download and extract the Jacquard Dataset.

Training

Training is done by the train_ggcnn.py script. Run train_ggcnn.py --help to see a full list of options, such as dataset augmentation and validation options.

Some basic examples:

# Train Squeeze on Cornell Dataset
python train_ggcnn.py --description training_example --network squeezenet --dataset cornell --dataset-path <Path To Dataset>

# Train Squeeze on Jacquard Datset
python train_ggcnn.py --description training_example2 --network squeezenet --dataset jacquard --dataset-path <Path To Dataset>

python -m utils.dataset_processing.generate_cornell_depth  ../cornell_grasp_data

python train_ggcnn.py --description training_example --network squeezenet --dataset cornell --dataset-path ../cornell_grasp_data

Trained models are saved in output/models by default, with the validation score appended.

Evaluation/Visualisation

Evaluation or visualisation of the trained networks are done using the eval_ggcnn.py script. Run eval_ggcnn.py --help for a full set of options.

Important flags are:

• --iou-eval to evaluate using the IoU between grasping rectangles metric.
• --jacquard-output to generate output files in the format required for simulated testing against the Jacquard dataset.
• --vis to plot the network output and predicted grasping rectangles.

For example:

python eval_ggcnn.py --network <Path to Trained Network> --dataset jacquard --dataset-path <Path to Dataset> --jacquard-output --iou-eval

python eval_ggcnn.py  --network squeeze_weights_cornell/epoch_00_iou_0.61_statedict.pt --dataset cornell --dataset-path ../cornell_grasp_data --iou-eval   --vis

Running on a Robot

Our ROS implementation for running the grasping system see https://github.com/dougsm/mvp_grasp.

The original implementation for running experiments on a Kinva Mico arm can be found in the repository https://github.com/dougsm/ggcnn_kinova_grasping.