A bio-inspired ventral–dorsal fusion module for enhancing visual representations
Deep neural networks have achieved remarkable success in visual recognition; however, learning stable and discriminative visual representations under diverse imaging conditions remains a persistent challenge. Existing approaches often rely on increasing model depth or complex attention mechanisms, which improve perform...