Learning Identity-Consistent Feature for Cross-Modality Person Re-Identification via Pixel and Feature Alignment

Figure 3

Channel attention submodule.