Detectron2 Object Detection
The notebook is intended to be used with google colab.
Training custom model using detectron2 takes some amount of tweaking.
- register custom dataset and metadata.
- add augmentations.
- add code to evaluate your model.
- create hooks to print validation loss and save best model.
This notebook does all of this, thus provides a good starting point for your project.
The notebook uses Wgisd dataset of grapes and retinanet as baseline model.
