InterstellarFang

> Hi > > I tried to inference the segmentation model(mask2former_beit_adapter_large_896_80k_cityscapes_ss) and got some weird point. > > (1) Run image_demo.py as CUDA_VISIBLE_DEVICES=1 python image_demo.py configs/cityscapes/mask2former_beit_adapter_large_896_80k_cityscapes_ss.py released/mask2former_beit_adapter_large_896_80k_cityscapes.pth.tar data/sample.png > >...

谢谢！那么想问下DcovN是不是描述CSMM模块具体参数的？

想问下common.py 里的convmixer函数的作用是什么，代表CSMM模块吗，谢谢！

你好，作者还在吗，我是从EMDloss那篇文章issue过来的,请问yolo等one-stage方法都不适用吗

然后方便提供邮箱吗，想进一步沟通算法细节

作者你好，我询问了EMDloss文章的原作者，得到的答复是yolo可以适用EMDloss

> Hey，guys！ In this PR: #210 After changing torch.repeat_interleave() to torch.expand(),, I successfully exported it under torch1.8.2+opset=12, But I'm not sure how this will affect performance. @UNeedCryDear Hello！Which python file...

> > @UNeedCryDear Hello！Which python file is this function (**torch.repeat_interleave()**) in,Can you tell me the location of this file？thanks a lot! > > https://github.com/facebookresearch/segment-anything/pull/210/files @UNeedCryDearThank you! I made changes based...

>  > > The code with a pink background has been replaced and you need to remove it. @UNeedCryDear I have added # to these two sentences with a...

> show me code you modefied. # Expand per-image data in batch direction to be per-mask # src = torch.repeat_interleave(image_embeddings, tokens.shape[0], dim=0) src_shape = (tokens.shape[0],*image_embeddings.shape[1:]) src = image_embeddings.expand(src_shape) src =...