GYxiaOH

请问下把binary map的loss换成dice loss是什么考虑？

4

RT...原文中是两个的cb loss，我尝试把binary换成cb发现在验证集精度一致是0，很奇怪，作者是实验后换的loss还是出于什么考虑呢

i find this code only can use roialign ,if i use roipooling ,Memory-Usage of GPU increases when training until out of memory. i add some new roi such as psroi...

when I use SN instead of BN, there is a big difference between val acc and train acc

1

when i use your code resnetv2sn50 ,it is normal difference between val and train(maybe ,i only run 10 epochs). but when i use my CNN model), there is a big...

你好，请问下你MNN的版本是啥啊，转mnn的时候有啥设置吗 我无论用你的onnx还是自己转的onnx都会提示Broad cast error Broad cast error, dim1 = 16, dim2 = 1600 Broad cast error, dim1 = 16, dim2 = 400 Broad cast error, dim1 = 16, dim2 =...

想问问，作者跑的时候有没有发现边界比较差

19

rt。。。发现最后的结果边界都比较差，会往内缩一点，不知道是不是我迭代的次数太少了，但是600个epoch也太长了吧，现在我一般只训50个epoch左右，600个估计一个星期能有一个结果

想问下，模型pretrain的时候用了那个类似UHD的切图吗？

6

如题。。如果pretrain就把图片切那么多份，训练成本是不是有些cover不住