TensorFlow: A Vegetable Classification System and Its Performance Evaluation

Recently in Japan, the aged population is increased and young generation is not interested in agriculture. For this reason, new systems and approaches based on Artificial Intelligence (AI) and Internet of Things (IoT) are needed. We focus on vegetable recognition using Deep Neural Network (DNN) as a method to visualize the tacit knowledge in the vegetable production. In our previous work, we evaluated a vegetable classification system using DNNs and proposed support tools for selling vegetables in the Internet. The vegetable classification system was based on Caffe and Chainer frameworks. In this paper, we present the performance of our vegetable classification consider TensorFlow framework using 10 kinds of vegetables. Also, we propose a VegeCare tool using AI, which manages the growth of vegetable for farmers. The evaluation results show that the learning accuracy was more than 70%. We found that the performance of vegetable classification results of leafy vegetables were degraded.