Announcement_11

🎉🎉🎉 After few round of review, our paper CAS-ViT has been accepted by IEEE Transactions on Image Processing(TIP’26). This method proposes a self-attention alternative structure based on convolution, which approximates self-attention without explicitly calculating the token-to-token relationship through a Sigmoid-based spatial and channel mechanism. CAS-ViT provides a feasible and effective alternative path for the structural design of ViT under resource-constrained conditions.