LiJiahang

Results 5 issues of LiJiahang

i ran the training process successfully,but i met two issues, first is that my mAP and losses are all 0: ![Screenshot from 2022-06-25 21-28-34](https://user-images.githubusercontent.com/103312600/175776037-ebf11932-83c1-45f5-893d-a1929b6ea41e.png) second is that i don't know...

i met an issue when i trained,this issue was like below: ![Screenshot from 2022-06-25 13-47-53](https://user-images.githubusercontent.com/103312600/175760413-44f8de8a-2345-4629-9b09-c48ed13ab994.png) why it is an invalid directory path? i just followed the README file,and i don't...

在训练yolov2自己的数据集时,eval产生的ap值非常低,只有0.003,已经训练了很多个epoch了但是还是不见好转,在之前用这个项目训练yolov1的时候就是正常的没有问题: - 如图 ![image](https://user-images.githubusercontent.com/103312600/182198604-5a99356c-7364-4d1f-a455-36a5b90eb1b8.png) 我训练时和yolov1相比就改了model为yolov2,为什么会出现这样的情况呢 是有什么地方没有定好参数么?望指正

作者您好: 您的工作非常优秀,我正在使用这个仓库进行二次开发。 但是我发现您的segmentation部分的BeiT实现和detection的BeiT实现有区别,在Attention类中,detection使用了窗口注意力,但是segmentation好像没有,您当时为什么要这样设计呢?是和task-specific有关吗?或者是有什么参考来源吗?我想研究学习一下,希望您能抽空回答一下我,再次感谢您的优秀工作!

I'm currently training Mask2Former implemented by MMseg, but I noticed that you have not support MaskFormer series models to calculate FPS and FLOPS, when could you fix it? I'm appreciate...