arXiv Computer Vision research summaries for January 17, 2024.

Today's Research Themes (AI-Generated):

• FAS models leverage real faces for improving generalization, achieving state-of-the-art cross-domain results.

• Hybrid CNN model with DiffStride and Spectral Pooling shows improved accuracy by maintaining image information.

• SAM, a vision foundation model, enables unsupervised change detection between optical imagery and map data.

• Cross-modality guidance-aided multi-modal learning enhances MRI brain tumor grading with dual attention.

• VideoCrafter2 overcomes data limitations for high-quality video diffusion models using synthesized images.