State-of-the-Art Deep Learning Architectures for Identifying Diseases in Cucumber and Black Gram Plants

Tackling the imperative task of plant disease identification, this paper proposes a holistic deep learning procedure applicable to cucumber and black gram crops. The study employs a diverse set of models, including the efficient Extreme Learning Machine (ELM), intricate Feedforward Neural Network (FNN), Deep Residual Network for hierarchical Convolutional Neural Network (CNN), feature extraction for spatial features, and the lightweight MobileNet, tailored for resource constraints.The study leverages the Cucumber and Black Gram Disease Classification Dataset, a comprehensive repository encompassing diverse disease stages. Employing a holistic methodology, the integrated models notably enhance the accuracy of disease classification. The Extreme Learning Machine enables swift learning, the Feedforward Neural Network captures intricate patterns, the Deep Residual Network explores hierarchical features, and the CNN excels in spatial feature extraction. Additionally, the lightweight MobileNet proves adaptable to resource constraints.In establishing a baseline, the collective performance of ELM, FNN, Deep Residual Network, CNN, and MobileNet was assessed. While MobileNet led with high classification accuracy at 97% for cucumber and 95% for black gram diseases, the baseline models collectively demonstrated commendable pathogen detection, maintaining an average precision of over 90% for both crops. MobileNet's superior accuracy sets a benchmark for the models, indicating its efficacy in disease classification.