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.
989
stars
201
forks
source link
非常感谢您的开源库,请问您修改backbone,有对应的预训练模型是怎么获得的? #65
Open
LeonNerd opened 2 years ago
为了实验,我想在您的库上做些修改或者添加新的backbone,所以想请教一下,修改backbone后获得对应的预训练权重的方法,可以通过v5的权重冻结获得么?还是只能通过重新训练