Early-Stage Turmeric Crop Disease Identification Using Multi-Scale Convolution and Transformer Attention

Le résumé fourni par la source

Early detection of diseases in turmeric plants is important because infections like leaf spot and rhizome rot spread fast and can lead to major losses in yield. While computer-vision techniques have improved plant disease diagnosis, older deep-learning models like ResNet and Inception still struggle with real field images and often miss the small, early-stage patterns that matter the most. To overcome this, the study uses MSCVT, a hybrid model that blends CNN layers for capturing fine textures with Vision Transformer attention for understanding global context and long-range relationships. Its multi-scale structure helps the model analyze symptoms at different resolutions, making it more effective at identifying faint spots, unusual textures, or early infection marks. Results show that MSCVT delivers clearer feature representation, more stable predictions, and higher accuracy than standard CNN-based models, making it a stronger option for real-world turmeric disease detection. This improvement can help farmers catch infections earlier, reduce crop loss, and make disease monitoring more efficient in practical agricultural settings.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.