Performance of Modified CNN Model on Rice Plant Diseases Detection

Identifying diseases in rice plants from images is an intriguing research area where techniques like ML and DL are applied from computer science area. The paddy plant diseases are classified by employing a model. The model was created through a comprehensive experimental analysis that employed deep learning and image processing techniques. Images of 13 distinct diseases in paddy crop including bacterial blight, leaf smut, tungro, leaf blast, and panicle blight are included in dataset. Each rice plant disease class consists of 1000 images. The entire dataset is the combination of images taken from Kaggle, Images of rice plants and a machine learning repository were obtained from the Poynad rice field in Alibaugh. The method evaluates three different segmentation techniques. Texture and color features are extracted, followed by applying K-means clustering to achieve best results. SVM is also implemented for multiclass classification. Dense-Net121, VGG16 and AlexNet models are trained and accuracy is compared with the Modified CNN model. The VGG16 model with powerful architecture represented accuracy of 99.44% of imagebased tasks. The AlexNet model demonstrated very efficient accuracy of 95.44%. The DenseNet121 model demonstrated maximum accuracy results of 99.89% with the efficiency in identifying different patterns in datasets. With depth analysis, it is observed that Modified CNN model achieved the best results by providing the accuracy of 99.91%.