GYxiaOH

...厉害了，那奇怪了，我把diceloss直接换成你定义的binary cross loss ，train上的acc最后是很高，但是test精度完全是0.。。作者对这个问题有头绪吗。。我看你第一个map用的也是binary cross loss

@WenmuZhou 。。那个模型被我覆盖掉了。。。看精度全为0的话我估计是。。。你有什么猜测吗

you can decrease learning rate(or init learnning rate in warm up) or dealy warm up iterations

@YanShuo1992 are you meet out of memory after some iterations? i meet same question , i compare psroi code with caffe2 and can't find some things.but i barely use CUDA...

@Whu-wxy 请问你调节到了多少哦

@Whu-wxy 好的，谢谢。另外我觉得train里那个图片没必要放大，因为validation的时候主要是看看结果好不好，不要求最高结果，train里图片要放大的话validation的速度就变慢了不少

@Whu-wxy python 3.7 重新编译pse在eval的时候会出现段错误，你遇到过这个问题吗

@Whu-wxy 我把decode的阈值调到0.58 f确实涨了，但是召回率掉了好多，这我没想通，理论上更多点了不应该召回率下降精确度升高吗。。。奇了怪了

I have same problem, do you have solutions? after 500-600 iterations , i can't continue my training because out of memory

crop roi pooling?I don't know what's your mean?can you explain in detail? i use psroi in my net and i compare psroi code and caffe codehttps://github.com/daijifeng001/caffe-rfcn/blob/4bcfcd104bb0b9f0862e127c71bd845ddf036f14/src/caffe/layers/psroi_pooling_layer.cu ,but i don't find...