Detectron2 is Facebooks new library that implements state-of-the-art object detection algorithm. This repository shows you how to use Detectron2 for both inference as well as using transfer learning to train on your own data-set.
See the official installation guide.
Using a pre-trained model for inference is as easy as loading in the cofiguration and weights and creating a predictor object. For a example check out Detectron2_inference_with_pre_trained_model.ipynb.
Training a model on a custom dataset is a bit more challenging because it requires that we register our dataset. For a example showing you how to train a balloon detection model check out Detectron2_train_on_a_custom_dataset.ipynb.
Gilbert Tanner