Comparative Analysis of Deep Learning Models for Accurate Detection of Rice Leaf Diseases

Detecting diseases in agriculture presents significant challenges, often stemming from improper treatment due to misdiagnoses of plant conditions. This study evaluates several deep learning models-VGG16, VGG19, ResNet50V2, and InceptionResNetV2-for the precise identification of various rice leaf diseases. Using a dataset of 2,628 images representing six distinct rice leaf diseases, the images are split into training (2,100 images) and validation (528 images) sets. Our findings show that InceptionResNetV2 achieves the highest training accuracy at 96%, while both ResNet50V2 and InceptionResNetV2 reach a peak validation accuracy of 97%. Notably, ResNet50V2 offers fewer model parameters (25.66 million) and a smaller memory size (97.92 MB), resulting in a shorter training time (91.97 minutes) compared to the other models, demonstrating its lightness without compromising performance.