What is WaterfowlDetector?

WaterfowlDetector is an end-to-end deep learning platform for developing and using artificial intelligence to detect and classify species of waterfowl on aerial color (RGB) images. We developed ten datasets from aerial images taken via crewed aircraft and a DJI Mavic Pro 2 uncrewed aerial systems (UAS, or drone). We currently have three detection models(Retinanet,Yolov5 and FasterRcnn) and three species classification models(Resnet, Resnext and Efficientnet). Instructions for training and executing your own custom models on your own bird datasets are also included.

Aerial color image taken by DJI Mavic 2 Pro at 60 meters with the detected birds and their confidence scores.