Methods › Computer Vision › Semantic Segmentation Modules › Channel-wise Cross Attention
Channel-wise Cross Attention
Introduced by Haonan Wang et al. in UCTransNet: Rethinking the Skip Connections in U-Net from a Channel-wise Perspective with Transformer
archive 2025-07-28 Description, source and code snippet are the archive's method entry.
Channel-wise Cross Attention is a module for semantic segmentation used in the UCTransNet architecture. It is used to fuse features of inconsistent semantics between the Channel Transformer and U-Net decoder. It guides the channel and information filtration of the Transformer features and eliminates the ambiguity with the decoder features.
Mathematically, we take the i-th level Transformer output 𝐎ᵢ ∈ℝ^(C×H×W) and i-th level decoder feature map 𝐃ᵢ ∈ℝ^(C×H×W) as the inputs of Channel-wise Cross Attention. Spatial squeeze is performed by a global average pooling (GAP) layer, producing vector 𝒢(𝐗) ∈ℝ^(C×1×1) with its $k$th channel 𝒢(𝐗) = 1/(H×W)∑ᴴᵢ₌₁∑ᵂⱼ₌₁𝐗ᵏ(i, j). We use this operation to embed the global spatial information and then generate the attention mask:
𝐌ᵢ = 𝐋₁ ·𝒢(𝐎ᵢ) + 𝐋₂ ·𝒢(𝐃ᵢ)
where 𝐋₁ ∈ℝ^(C×C) and 𝐋₂ ∈ℝ^(C×C) and being weights of two Linear layers and the ReLU operator δ(·). This operation in the equation above encodes the channel-wise dependencies. Following ECA-Net which empirically showed avoiding dimensionality reduction is important for learning channel attention, the authors use a single Linear layer and sigmoid function to build the channel attention map. The resultant vector is used to recalibrate or excite 𝐎ᵢ to 𝐎̅ᵢ = σ(𝐌ᵢ) ·𝐎ᵢ, where the activation σ(𝐌ᵢ) indicates the importance of each channel. Finally, the masked 𝐎̅ᵢ is concatenated with the up-sampled features of the i-th level decoder.
Papers archive 2025-07-28
5 shown of 5, newest first. Repository counts are the archive's code-links table. A Syntology line states what Syntology ran from that paper's harvested code; it is per sample and not a correctness claim.
-
Enhancing Conditional Image Generation with Explainable Latent Space Manipulation 29 Aug 2024 · 1 repository · arXiv:2408.16232
-
Boosting Medical Image Segmentation Performance with Adaptive Convolution Layer 17 Apr 2024 · 1 repository · arXiv:2404.11361
-
ACC-UNet: A Completely Convolutional UNet model for the 2020s 25 Aug 2023 · 1 repository · arXiv:2308.13680
-
LViT: Language meets Vision Transformer in Medical Image Segmentation 29 Jun 2022 · 1 repository · arXiv:2206.14718Syntology ran 1 of 1 samples · 0 unverified
-
UCTransNet: Rethinking the Skip Connections in U-Net from a Channel-wise Perspective with Transformer 9 Sep 2021 · 3 repositories · arXiv:2109.04335Syntology ran 1 of 1 samples · 0 unverified · 1 pointer-only (licence)
Tasks archive 2025-07-28
11 tasks the archive attaches to papers tagged with this method, by distinct papers. A task without a page in the catalog is plain text.
Usage over time archive 2025-07-28
Components: the archive holds no method-to-method composition, so PwC's Components table cannot be rebuilt; the Papers list carries no Results column for the same reason (the archive does not join its leaderboard rows to method tags).
Categories archive 2025-07-28
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections