Abstract: Highlights•Channel-wise Cross Fusion Transformer; Channel-wise Cross Attention.•Faster convergence and higher robustness in the extraction of both low-level and high-level features.•Solving the semantic gap between the encoder and the decoder.•Fuse the two networks via CRFB to fully exploit the advantages of both networks.
External IDs:dblp:journals/eaai/JiangLH22
Loading