YOLOV5_NCNN_Android
YOLOV5_NCNN_Android copied to clipboard
YOLOv5 C++ Implementation on Android using NCNN framework
将自己的模型转为ncnn,inference结果多了好多误检
你好,我解决了自己训练的模型,部署时预测框都在左上角不对的问题了。 【1】在导出onnx模型时,common.py删除切片操作还是采用的return self.conv(torch.cat([x,x,x,x], 1))替代。然后export.py导出模型时,输入大小改为训练模型时的一半,如输入大小640时为[320, 320]   【2】预测框不对是因为YoloV5.h文件中,预测层没有根据自己模型指定对,原来的是“394,“375,“output”这3层输出,可以根据Netron查看自己模型的3个预测层修改,我的是“output”,“423”,“442”,然后预测框正确。   感谢之前各位大佬的回答和分享! _Originally posted by @Skr20200701 in https://github.com/sunnyden/YOLOV5_NCNN_Android/issues/3#issuecomment-668028381_
The result image of detection has the same labels and an small box in the up left of the image, why?
I have use my yolov5 model to replace yours, the process is that *.pt -> *.onnx -> *-sim.onnx -> *-sim.param, *-sim.bin. Then use the Android Studio to Synic..., Build ->...
请问一下,我用onnx2ncnn转换yolov5s.onnx的时候出现Segmentation fault (core dumped)这种错误,怎么处理?我用的命令是 ./onnx2ncnn yolov5s-sim.onnx y5.param y5.bin。这种怎么解决?
Hi, Thanks for you work i using my own trained model file will report error Slice operation in .pt or in .onnx?