An Improved System for Detection of Tea Leaf Disease Using Hybrid AI Models

The Agricultural domain is segmented into various categories, such as grains, fruits, vegetables, and flowers. Leaf diseases are commonly found in all of these categories. Deep learning (DL) and Machine learning (ML) models are used to classify diseases and investigate the most recent advancements to identify plant leaf diseases. This paper presents three hybrid deep learning models with the anticipation of improved results. Three separate pre-trained deep learning models have been utilized and a hybrid approach is proposed. The pre-trained modes like CNN, VGG19, and ResNet50 are merged using transfer learning and global average pooling features to create three hybrid models which are the combination of VGG19 with global average pooling, VGG19, ResNet50, and CNN with the use of transfer learning and the third model is the merging of VGG19 with features of CNN model. The proposed models use transfer learning and global average pooling concepts to yield improved results. The three hybrid models are Model 1, Model 2 and Model 3. In this study, we have investigated these models in terms of their accuracy in diagnosing diseases that affect tea leaves. Various images of healthy and diseased tea leaves in the dataset were used to train and test the models. Model 3 achieved a 99.65% accuracy making it the most successful whereas Model 2 and the Model 1 models achieved 98.7% and 98.58% accuracy respectively using 5-fold cross-validation. Our research indicates that these models can detect tea diseases accurately and efficiently, potentially enhancing early disease identification and crop quality. Professionals can assist quickly and easily in detecting diseases in plant leaves early.