Investigating the Impact of Dataset Size on Model Complexity for Leaf Disease Recognition using CNN

Detecting plant diseases poses a formidable challenge in agriculture, impacting crop yield and environmental sustainability. Timely and accurate disease prediction can enhance crop production and market value while minimizing ecological harm. The effectiveness of Deep learning models in recognizing plant leaf diseases depends on the quality and quantity of the training data and the model's architecture. Large Datasets present significant challenges regarding memory requirements, Computation, and storage. This can be solved by developing Deep Learning models that perform well with limited training data. This work comprehensively explores approaches for developing efficient deep learning models that can perform well with limited training data. The work considers two scenarios to test the system's performance: limited data with a comparatively complex CNN model and comparatively larger data with a simple model; both are good at accuracy, but deployment model design plays a role. Thus the proposed algorithm can help the model converge more efficiently, making the most of the available data and potentially mitigating the impact of data scarcity.