ruiming46zrm

关于yolov5的移植问题

1

我在作者的基础上修改一些代码，移植好了yolov5，但目前有些问题没搞明白： 用high-precision量化的wk跑起来仍然有一些精度损失，主要表现为confidence略有下降，对于本身得分就不高的目标容易发生漏检，请问这个现象是合理吗？还是说哪里弄得有问题，现在我也在检查很久了，没发现问题。 ps：(最好的办法是每一层打印输出和torch模型对比，但可能我太菜了。。觉得这个工作量有点难受)

maybe few differences with mxnet

7

Hi, @TreB1eN , I'm learning your codes for few days ,it's very nice work, but compared the network with mxnet version: 2 key differences found as following: 1 . in...

in model , Arcface. self.mm = self.sin_m * m # issue 1 we know it linked to cosface but why margin is sin_m*m.? and what's the issue 1 ? thankyou

can not perform same as your pretrained res101

13

hello Evaluation: LFW Acc: 0.9968333333333332, CFP_FP Acc: 0.8959999999999999, AgeDB Acc: 0.9726666666666667 it's my training result as 20 epoch , hyper param as you set .. acc on cfp-fp is quite...

精度损失

2

请问用高精量化后仍有一定精度损失是否正常？ 平均confidence会掉2-3个点， 偶尔也掉10个点+ ， 会漏检

 how can i solve this problem ? it seems somthing wrong . i tried resnet50 and resnet50x64, it's the same result. my command is : ` python demo/demo.py --input...