Feature Extraction In YOLOV4

Question

I know the feature extractor in YOLOv4 is CSPDarknet-53, and it has been pre-trained on the ImageNet. But I\'m wondering how exactly the feature extraction process is done.<