luna9806 comments

Results 4 comments of


                                            luna9806

the precision of detection

> In the process of my reproduce, the precision of detection is relatively low。Is there some problem？ Looking forward to your reply！Thanks! Do you know why now?

RuntimeError: weight tensor should be defined either for all or no classes

I have the same problem : Starting training for 100 epochs... Epoch gpu_mem box obj cls labels img_size 0%| | 0/63 [00:01

RuntimeError: weight tensor should be defined either for all or no classes

> @midasklr @yanggui19891007 , 这个bug具体是因为，作者在读取分割的label的时候，错误的使用了灰度图像，然而VOC数据集恰好有两个颜色的灰度值是一样的，所以没有区分背景以及边缘，以至于20可以跑通，而换了一个数字之后就不可以了，另外的证据是show最终效果的时候，整个图片都盖上了mask，再次证明作者没有区分背景，希望作者可以修改一下Code！ > > 作者的其他Code还是写的非常不错的，希望能把这个小Bug尽快修改一下：）请问代码您修改了吗？可以说说怎么修改吗？

RuntimeError: weight tensor should be defined either for all or no classes

@acekrystal @d0coco @yanggui19891007 @mahadayoub @XJTUNR 请问“RuntimeError: weight tensor should be defined either for all or no classes ”你们解决了吗？代码怎么改的可以说说吗? Is this problem solved? How was the code modified?