In this repository, I use multiple techniques to improve the performance of a single class object detection(Kangaroo) with a small dataset.
Using YOLOv4 from https://github.com/miemie2013/Keras-YOLOv4/ repository.
Also inspired from https://github.com/matterport/Mask_RCNN/blob/master/mrcnn/utils.py for mAP calculations
Working Notebook (Final) : KangarooDetection.ipynb
Notebook for future developement: Copy_of_KangarooDetection.ipynb