Design and Development of Tomato Leaf Classification Framework Using CNN

Crop production loss in agriculture is twenty-two percent because of biotic and abiotic stress/disease and living and non-living organisms. Producers must be able to identify these requirements from the beginning. Computer vision can help discover early illness patterns and clusters. Recent Deep learning-based computer vision advancements can recognise single or multi-biotic stress in tomato plant leaves. This investigation will use the PlantVillage dataset to segment objects. The model was trained with tagged, enriched, and augmented data. The Hybrid-CNN model was used to segment sick tomato plants. In object detection, the proposed Hybrid CNN outperforms the U-Net architecture and Seg-Net when paired with instance segmentation. This study used semantic segmentation data to identify and classify single-and multiple-leaf illnesses. The Hybrid-CNN algorithm's predicted output corresponds to that of the adapted U-Net, M -SegN et, and adapted U -Net models based on state-of-art evaluation parameters with other evaluation parameters such as Intersection over Union (IoU) and Mean IoU. The proposed model outperformed competitors, processing 1004 pictures in under 30 nanoseconds. With an accuracy of 98.580/0, the suggested model outperforms modified segmentation models.