ffhh

Results 46 comments of ffhh

路径还是别用加号加来加去的吧

@1093842024 I test the script on CPU(1 core), the inference time is about 70ms.

So, any one uses the MobileNetV3 as backbone?

@xsacha you mean the speed is lower than v1 on GPU? And if I use the mobilenet backbone, I would use the retinaface on edge devices.

@rydenisbak Can you share the results of mobilev2 on the widerface?

please look at this line: pos1 = conf_t > zeros ------------------ 原始邮件 ------------------ 发件人: "zhangyuan1994511"

https://github.com/biubug6/Pytorch_Retinaface/issues/10#issuecomment-567029509

I just discard the batch like this..., and have not made in-depth analysis to it.

I experimented with PAnet, MnasFPN and ASFF. From the test results, MnasFPN is the best among these models, but size of the weight increase 0.9M than FPN.