本项目是在 B站Up主 哈萨克斯坦 教程下，结合自己的实验室硬件，修改调试完成的

哔站demo视频：https://www.bilibili.com/video/BV11m4y1w7wH/#reply135077881808

## Requirements
- numpy
- opencv-python
- matplotlib
- scikit-image
- imageio
- torch
- torchvision
- torchsummary
- tensorboardX
- pyrealsense2
- Pillow

#### Cornell Grasping Dataset

1. Download and extract the cornell Dataset
链接：https://pan.baidu.com/s/1rHz-79Mt47Dv_od-7uuEKg 提取码：8888
2. Convert the PCD files to depth images by running `python -m utils.dataset_processing.generate_cornell_depth <Path To Dataset>`

#### Jacquard Dataset

1. Download and extract the Jacquard Dataset
链接：https://pan.baidu.com/s/1524HrVAoHNlc6-9lcZaGew 提取码：8888

## Model Training

A model can be trained using the `train_network.py` script.  Run `train_network.py --help` to see a full list of options.

Example for Jacquard dataset:

```bash
python train_network.py - --dataset-path <Path To Dataset> --description training_jacquard 

Model Evaluation

The trained network can be evaluated using the evaluate.py script. Run evaluate.py --help for a full set of options.

Example for Jacquard dataset:

```bash
python evaluate.py --network "trained-models/jacquard-d-grconvnet3-drop0-ch32/epoch_50_iou_0.94" --iou-eval --input-size 300 --use-rgb 0

python evaluate.py --network "trained-models/jacquard-rgbd-grconvnet3-drop0-ch32/epoch_42_iou_0.93" --iou-eval --input-size 300

Run Tasks

A task can be executed using the relevant run script. All task scripts are named as run_<task name>.py. For example, to run the grasp generator run:

python plane_grasp_real.py

Run on a Robot

python run_plane_grasp_real.py

补充

本项目还需要下载一下ur-rtde

https://gitlab.com/sdurobotics/ur_rtde