This code follows the implementation architecture of Detectron2, https://github.com/facebookresearch/detectron2.
The goal of this project is to find which objects are in the images and also to detect its locations.
Run train.py
In train.py, register my own custom dataset which is in the coco format. Then, I applied Imagenet pretrained weights to train the model.
Run inference.py
In inference.py, evaluate the training results on the weights that I trained. Additionally, visualize the training results on images and save them in the output_dir.