Triple-Head Spatial Attention-Based Deep Learning Approach for Plant Disease Identification

Global food security is seriously threatened by plant diseases, which calls for creative approaches to early diagnosis and efficient treatment. Computer vision techniques provide scalable alternatives to labor-intensive and in-accurate traditional diagnostic procedures. This research focuses to upgrade the diagnosis of plant diseases by integrating attention mechanism with deep convolutional model. More specifically, we propose triple-head spatial attention module which integrated to EfficientNet-B0 for enhancing feature representation ability. The PlantDoc dataset, which includes images from 27 classes, was used to assess the performance of several model variations. With an accuracy of 62.59% for uncropped and 83.97% for the cropped dataset, the presented model showed excellent feature extraction and classification compared to prior methods.