Defectron

Defect detection framework on cylindrical metal components on robotic cell using computer vision technologies (deep learning models).

Dependencies

Torch 1.10.1

pip install torch==1.10.1+cpu torchvision==0.11.2+cpu torchaudio==0.10.1 -f https://download.pytorch.org/whl/cpu/torch_stable.html

Install CoppeliaSim Edu from here
Install all dependencies above using the following command:
```
pip install -r requirements.txt
```
Note you also need to install yolov7 requirements in the yolov7 folder.
Install zmqRemoteApi (CoppeliaSim API) from here
Download supporting files and models from here and put YoloV7 model folder inside the src folder.

Warning Don't change the folder name of the YoloV7 model folder. It should look like this src/yolov7_model.

To run this program make sure you installed all dependencies and supporting files and models.
Run CoppeliaSim Edu and open the scene file scene.ttt in the data/cell folder.

Note Don't start the scene from CoppeliaSim, it will be started automatically from the program.
Run the main entry points which is defectron.py file in the main folder.
Make sure you change the workpiece information in the defectron.py file to match your workpiece, if you are using a new workpiece different from the one in the scene.
When you run the defectron.py file and the scene is started, make sure you speed up the simulation to its maximum speed by pressing the "bunny" button in the top right corner of the CoppeliaSim window.
Now, relax and enjoy the show. 😄

Note If the run fails from the first time, try to run it again as YoloV7 might require to download the onnx package.

sample1_trimmed.mp4

sample2.mp4