We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
找了一些原因,可能是由于在原版yolov5中: # Compute conf x[:, 5:] *= x[:, 4:5] # conf = obj_conf * cls_conf 但是我发现我这边得到的obj_conf(detection[5])比原版的要低很多,乘下来就只有0.2左右了。 我不是太清楚,这个是怎么计算的,想请教一下您😂 在您的代码中也有类似的描述: ###其实只需要对x,y,w,h做sigmoid变换的, 不过全做sigmoid变换对结果影响不大,因为sigmoid是单调递增函数,那么就不影响类别置信度的排序关系,因此不影响后面的NMS ###不过设断点查看类别置信度,都是负数,看来有必要做sigmoid变换把概率值强行拉回到0到1的区间内
谢谢!
The text was updated successfully, but these errors were encountered:
我的也是同样的情况,置信度都接近于1了。这是怎么回事呀
Sorry, something went wrong.
原版的yolov5里,置信度等于目标框置信度和类别置信度的乘积,我这里的程序没有做这一步处理,导致最后得到的置信度很高,你可以在程序里加上这一步
你是说在做归一化之前吗?这里应该单独标出类别置信度吧,还是标记目标框置信度和类别置信度?
No branches or pull requests
找了一些原因,可能是由于在原版yolov5中:
# Compute conf
x[:, 5:] *= x[:, 4:5] # conf = obj_conf * cls_conf
但是我发现我这边得到的obj_conf(detection[5])比原版的要低很多,乘下来就只有0.2左右了。
我不是太清楚,这个是怎么计算的,想请教一下您😂
在您的代码中也有类似的描述:
###其实只需要对x,y,w,h做sigmoid变换的, 不过全做sigmoid变换对结果影响不大,因为sigmoid是单调递增函数,那么就不影响类别置信度的排序关系,因此不影响后面的NMS
###不过设断点查看类别置信度,都是负数,看来有必要做sigmoid变换把概率值强行拉回到0到1的区间内
谢谢!
The text was updated successfully, but these errors were encountered: