ViTaL: An Advanced Framework for Automated Plant Disease Identification in Leaf Images Using Vision Transformers and Linear Projection for Feature Reduction

Our paper presents a robust framework for automated plant disease identification in leaf images, incorporating key stages for enhanced accuracy. The pre-processing phase employs thumbnail resizing and normalization for computational efficiency without losing critical details. Feature extraction utilizes Vision Transformers, with variations exploring linear projections. Various Convolutional Neural Network (CNN) architectures assess the impact of linear projection on performance metrics. The top-performing model achieves a Hamming loss of 0.054. A novel hardware design using a Raspberry Pi Compute Module addresses low-memory configurations, ensuring practicality and affordability. This research contributes valuable insights and tools for early plant disease detection, potentially improving crop yields and food security.