Tomato leaf disease detection using convolutional neural network

Abstract

The fertile soil and easy access to water make agriculture more suitable and valu able for Bangladesh. Most people are directly or indirectly dependent on agricultural products for their livelihood. Agriculture plays an important role in the GDP of Bangladesh, which is 12.68% in 2019. According to the UN FAO, tomato is a type of vegetable that is ingested by 16% of the entire population. When analyzing the agricultural environment in Bangladesh, tomatoes are considered one of the most common vegetables. Plant infections pose a significant danger to crop production, yet timely detection remains a challenge in several regions of the world due to a lack of facilities. Climate changes are forcing us to take more care of agriculture to ensure food safety. Early detection of diseases has been made possible by cur rent developments in computer vision. Image processing and deep learning are very useful in this situation. The object’s impacted region is segmented using a bespoke threshold algorithm based on HBS (hue-based segmentation). Utilizing a color co occurrence approach, the segmented portion’s consequential selected features are recovered for edge detection. This research shows the diagnosis and detection of tomato leaf diseases involving several steps, including image capture, image pre processing, picture segmentation, feature extraction, and classification using a Con volutional Neural Network(CNN). The proposed CNN model achieved 95% accuracy while using much fewer computational resources, which makes it easily deployable in mobile applications.