SLCAM-AquaNet: an attention-enhanced lightweight deep learning model for accurate classification of aquaculture disease

Timely and accurate classification of bacterial, viral, parasitic, and fungal diseases in aquaculture is essential for sustaining fish health, optimizing yield, and securing the global food supply. Aquatic diseases significantly compromise productivity, quality, and economic viability in aquaculture systems, thereby posing a serious threat to food security amid rising global demand. Early detection and effective disease management are therefore, critical to minimizing losses, improving yield, and curbing the spread of infections. In this study, we propose a novel lightweight Spatial Location Channel Attention Module (SLCAM) to facilitate accurate identification and classification of aquaculture diseases affecting key farmed species, including Tilapia, Grass carp, Catla, Rohu, Nile perch, and Channel catfish. The SLCAM leverages a four-branch architecture to compute attention weights across horizontal, vertical, spatial, and channel dimensions, enabling adaptive feature refinement. By integrating spatial and locational awareness with channel-wise attention, the model enhances representational capability while maintaining computational efficiency. Experimental results demonstrate that proposed system achieves high classification accuracies: 96.89% for Tilapia, 96.72% for Grass carp, 97.00% for Catla, 97.78% for Rohu, 96.50% for Nile perch, and 97.08% for Channel catfish. Moreover, SLCAM-integrated models exhibit significant Top-1 accuracy improvements of up to 5.78%, underscoring their effectiveness for real-time, resource- efficient disease monitoring in aquaculture environments.