This repository provides a comprehensive suite of model compression techniques for YOLOv5 object detection models. It includes multi-backbone substitutions using architectures like Ghostnet, ShuffleNetv2, Swin Transformer, and MobileNetv3. The toolbox implements pruning methods (EagleEye, Network Slimming) and quantization via MQBench, and supports deployment optimization through TensorRT and ncnn runtimes for edge device inference.