This repository provides a complete implementation of the YOLO3 object detection algorithm using Keras and TensorFlow. It supports full training pipelines including warmup and multi-scale training, multi-GPU acceleration, and evaluation using mean Average Precision (mAP). The project includes pre-trained weights and configurations for detecting various object classes across custom and standard datasets like VOC and raccoon imagery.