Publication

Abstract : Vegetable detection and classification is a challenging objective in daily production and use, and the complexity increases when other parameters such as shape, size and color are taken into consideration. In this paper, we define a methodology that will detect and classify three different green vegetables of different sizes. This method involves the use of Tensor Flow, Dark flow which is a Tensor flow version of You only look once (YOLO) algorithm, OpenCV. To train the desired network, several various vegetable images were fed into the network. The images were pre-processed before training by drawing bounding boxes around the vegetable manually using OpenCV. The main algorithm responsible for the detection and classification is YOLO. This method provides a faster and smarter way to identify an object in the given image or video. Once the network is trained, the test input is passed in to the network. Once the input is provided to the network, the output will display bounding boxes around the recognized vegetable and label it with its predicted class category with accuracy of 61.6 percent.