YOLOv5 Series Multi-backbone(TPH-YOLOv5, Ghostnet, ShuffleNetv2, Mobilenetv3Small, EfficientNetLite, PP-LCNet, SwinTransformer YOLO), Module(CBAM, DCN), Pruning (EagleEye, Network Slimming), Quantization (MQBench) and Deployment (TensorRT, ncnn) Compression Tool Box.
994
stars
200
forks
source link
我将yolov5修改为加c2o的格式，网络结构和训练参数完全没变化，但是性能大幅下降，有大佬知道为什么吗