This repository provides a PyTorch implementation of YOLOv4, a state-of-the-art object detection model. The implementation supports object detection, instance segmentation, and semantic segmentation tasks. It includes various advanced loss functions (RS loss, aLRP loss, AP loss, alpha IoU), attention modules, and training techniques. The project is based on the original Darknet implementation and Ultralytics’ YOLOv3 work.