Ruoyu Chen

> > @zht-ttt 可以用backbone的最后一个卷积层做，实现上更简单，参考分类的Grad-CAM即可 > > 你好，能讲讲具体如何修改吗？我是小白，目前还在学习这一块 demo里面的layer name自定义改一下

> 是不是把网络换成我的就可以用了。。 对于有FPN的网络，你需要自己指定last layer，可以选`roi_heads.box_pooler`

> What changes have you made when you applied it on the FPN model? @hanikh As she mentioned, you can't use the layer name judged by function `get_last_conv_name`, you need...

> I use the edl loss to train in mini-imagenet dataset with 64 classes, but the loss can't converge and the accuracy is very low. I think you can modify...

> > > I use the edl loss to train in mini-imagenet dataset with 64 classes, but the loss can't converge and the accuracy is very low. > > >...

> > > > > I use the edl loss to train in mini-imagenet dataset with 64 classes, but the loss can't converge and the accuracy is very low. >...

> > Hi, > > Thanks a lot for your pytorch version code. > > I am training my own dataset with this model. But, the intermediate file till 80...

> > 请问，是把rio_heads.py里面所有的storage = get_event_storage()都改成with EventStorage(): storage = get_event_storage()吗？ > > 是都改掉吗 凡是报错地方都改掉，我改了好几个文件，像`rcnn.py`什么的，报错就改： ```python from fsdet.utils.events import get_event_storage, EventStorage ... with EventStorage(): storage = get_event_storage() ``` 另外我觉得builder这个fsdet太麻烦了，直接`pip install detectron2==0.4`（这个要用官方链接）多好，免得出现了builder时候有错误，另外还是建议pytorch 1.4或者cuda不兼容情况。

> > Hello, how did you compile the code on the 3090 using cuda 11.0. > > I can not compile the code on the 3090, but you can run...

> > Please refer to #3 > > I am sorry to take up your time.Did you get through the t-sne？which file that code should be added to？Thank you very...