为什么我的loss负几百，有一样情况的吗，我没有更改代码

[RyougiShiki12]

我也是训练的有这个问题，请问你解决了没有

[RyougiShiki12]

尴尬，我也是感觉差别有点大

[Luiweony]

我也是训练有这个问题，而且效果还不怎么好，请问你们解决了吗？是不是那里没有修改正确？

[RyougiShiki12]

我也是训练有这个问题，而且效果还不怎么好，请问你们解决了吗？是不是那里没有修改正确？

我感觉他代码有问题。。我吧CDD放上去直接RUN的

[lehaifeng]

我感觉他代码有问题。。我吧CDD放上去直接RUN的

please refer to readme file. The main reason might be that you did not revise the label from the original label value 0-255 to the network value 0-1.

[coocooon]

我感觉他代码有问题。。我吧CDD放上去直接RUN的

please refer to readme file. The main reason might be that you did not revise the label from the original label value 0-255 to the network value 0-1.

Hello, may I ask how to convert the pixel value to 0 to 1. Is it to directly divide the pixel value by 255, or first turn it into a grayscale format, and then divide by 255.Thank you very much and look forward to your reply.

[yiyi-today]

我感觉他代码有问题。。我吧CDD放上去直接RUN的

please refer to readme file. The main reason might be that you did not revise the label from the original label value 0-255 to the network value 0-1.

你好，请问怎么把原始label的值从0-255转换成0-1，谢谢！

[dmzlingyin]

情况就是这么个情况，具体咋改也不知道

Epoch [27/1120] Loss: -752.6592 Mask_Loss_conv5: -251.0686 Mask_Loss_fc: -250.4772 Mask_Loss_embedding: -251.1134 Epoch [27/1140] Loss: -673.4584 Mask_Loss_conv5: -224.4770 Mask_Loss_fc: -224.6068 Mask_Loss_embedding: -224.3746 Epoch [27/1160] Loss: -148.1745 Mask_Loss_conv5: -49.3769 Mask_Loss_fc: -49.4018 Mask_Loss_embedding: -49.3958 Epoch [27/1180] Loss: -71.9143 Mask_Loss_conv5: -23.9587 Mask_Loss_fc: -24.0065 Mask_Loss_embedding: -23.9490 Epoch [27/1200] Loss: -362.3986 Mask_Loss_conv5: -120.7738 Mask_Loss_fc: -120.8340 Mask_Loss_embedding: -120.7908 Epoch [27/1220] Loss: -190.4212 Mask_Loss_conv5: -63.4736 Mask_Loss_fc: -63.4761 Mask_Loss_embedding: -63.4716 Epoch [27/1240] Loss: -55.3082 Mask_Loss_conv5: -18.4365 Mask_Loss_fc: -18.4353 Mask_Loss_embedding: -18.4363 0.058823529411764705 0.0030097260885289996 Epoch [28/0] Loss: -342.8940 Mask_Loss_conv5: -114.2997 Mask_Loss_fc: -114.3042 Mask_Loss_embedding: -114.2901

[li670]

兄弟解决了吗

[stiwfmjX]

我感觉他代码有问题。。我吧CDD放上去直接RUN的

please refer to readme file. The main reason might be that you did not revise the label from the original label value 0-255 to the network value 0-1.

你好，原始CDD数据集label的值是在0到255之间的，请问怎么将这个灰度图转换成为二值图，介于0-255之间的数要怎么处理呢？

[1998wq]

我也是训练有这个问题,而且效果还不怎么好,请问你们解决了吗？是不是那里没有修改正确？

你好，我想问一下，这数据集的标签不就是二值图像，就是0和1吗，怎么要转换

[1998wq]

情况就是这么个情况,具体咋改也不知道

epoch[27/1120]Loss:-752.6592 Mask Loss con V5:-251.0686 Mask Loss fc:-250.4772 Mask Loss embedding:-251.1134epoch[27/1140]Loss:-673.4584 Mask Loss con V5:-224.4770 Mask Loss fc:-224.6068 Mask Loss embedding:-224.3746epoch[27/1160]Loss:-148.1745 Mask Loss con V5:-49.3769 Mask Loss fc:-49.4018 Mask Loss embedding:-49.3958epoch[27/1180]Loss:-71.9143 Mask Loss con V5:-23.9587 Mask Loss fc:-24.0065 Mask Loss embedding:-23.9490epoch[27/1200]Loss:-362.3986 Mask Loss con V5:-120.7738 Mask Loss fc:-120.8340 Mask Loss embedding:-120.7908epoch[27/1220]Loss:-190.4212 Mask Loss con V5:-63.4736 Mask Loss fc:-63.4761 Mask Loss embedding:-63.4716epoch[27/1240]Loss:-55.3082 Mask Loss con V5:-18.4365 Mask Loss fc:-18.4353 Mask Loss embedding:-18.4363 0.058823529411764705 0.0030097260885289996epoch[28/0]Loss:-342.8940 Mask Loss con V5:-114.2997 Mask Loss fc:-114.3042 Mask Loss embedding:-114.2901

你好，我也出现了这个情况，可以添加你联系方式探讨一下这个问题吗 我qq:1994462220 祝你好运

[WesternTrail]

我是用的BCD跑的，这命令行看的我强迫症都犯了，epoch开始那是什么鬼？还有怎么还有负数的loss???