A Novel Triple Attention-Based Deep Learning Framework for Accurate Pomegranate Disease Detection

Abstract

Pomegranate disease detection is critical for ensuring crop quality and productivity. This research proposes a novel deep learning framework that leverages triple attention mechanisms and depth-wise separable convolutions to accurately identify pomegranate diseases. The framework incorporates a pre-processing stage using Savitzky–Golay filtering and CLAHE for noise reduction and contrast enhancement. Quantum-based Sobel edge detection is employed for feature extraction, followed by adaptive sunflower optimisation for feature selection. The TAtt-DSC model, optimised with the CBRCM algorithm, effectively classifies healthy and unhealthy pomegranate fruits. Experimental results demonstrate superior performance with precision of 97.2%, recall of 93%, accuracy of 99.14% and F1-score of 95.5%. This innovative approach offers a promising solution for efficient and accurate pomegranate disease diagnosis.