zhangjinsong3

> 如果没有使用阈值图，那是因为什么呢？是因为用了阈值图效果没有以前好，还是都差不多呢？ @ @WenmuZhou 论文作者说了, 在推理的时候,用不用thresh map,效果差不多,为了体现速度优势,推理的时候就不加thresh map了.,这样可以少跑一个head

It seems that 'torch.linspace' is not supportted in onnx, After bypassing it, torch.onnx.export may work. Not tested yet!

> The problem is solved. Besides following the code of [https://github.com/lbin/DCNv2](url), I add the following three lines before `return` in the _backward function of _DCNv2 in dcn_v2.py: > > ```...

> Thanks for your excellent work! > > I use this command to train net: python [EXP_PATH] --amp_backend native -b 8 --gpus 8 But only gpu 0 works. How to...

Thanks for your amazing work! And how about the speed in cpu? As I know, opencv 3.4.2 has supported yolov3 inference with the speed of 220 ms on an 12x...