Spatio-Channel Attention Blocks for Cross-modal Crowd Counting

AbstractCrowd counting research has made significant advancements in real-world applications, but it remains a formidable challenge cross-modal settings. Most existing methods rely solely on the optical features of RGB images, ignoring feasibility other modalities such as thermal and depth images. The inherently differences between different diversity design choices for model architectures make crowd more challenging. In this paper, we propose Cross-modal Spatio-Channel Attention (CSCA) blocks, which can be easily integrated into any modality-specific architecture. CSCA blocks first spatially capture global functional correlations among multi-modality with less overhead through spatial-wise attention. spatial attention are subsequently refined adaptive channel-wise feature aggregation. our experiments, proposed block consistently shows performance improvement across various backbone networks, resulting state-of-the-art results RGB-T RGB-D counting.KeywordsCrowd countingCross-modalAttention