VW
VW copied to clipboard
iclr2024 poster Varying Window Attention
Can you release the weights of cityscapes? I just find the weights of ade20k in your link. thanks!
哈喽,请问能提供一下SegFormer-B0的config文件吗,因为我按论文里的说法配置后效果不好,可能是我没写好配置文件的原因。 > The new flow of channels is 128 (output of multi-layer aggregation)→ 512(concatenation of learned multi-scale representations)→ 128(output of multiscale aggregation)→ 128 + 32 = 160(concatenation of LLE)→ 128(final...
Hi, In your first draft of paper on Arxiv, you mentioned that you are using MLP mixer to mix the channels but I don't see any code that uses MLP...
HuggingFace -> Hugging Face