Protein-Based Fish Species Identification: Dataset, Models, and Insights from Native Bangladeshi Fish
Researchers have developed a new method for identifying fish species in Bangladesh using protein sequences. They created a dataset of protein sequences for nine native Bangladeshi fish species and evaluated several machine learning architectures. Their proposed hybrid model, MotifCNN-Transformer+TA-PE, achieved 79.80% accuracy and is significantly faster and smaller than the fine-tuned ProtBERT model, making it more practical for resource-constrained environments. AI
IMPACT This research could improve fisheries management and food security in regions with limited computational resources.