วารสารวิชาการเทคโนโลยีอุตสาหกรรม (J. Ind. Tech.) อยู่ในฐานข้อมูล TCI กลุ่ม 2 (2563) มีค่า JIF = 0.094 | The Journal of Industrial Technology (J. Ind. Tech.) is indexed in TCI Tier 2 (2020) with impact factor, JTIF 0.094

A System for Cooking Recipe Sharing and Cooking Recipe Finding by an Image of Ingredients using Deep Learning Technique
Type Of : Academic Manuscript
Type : Electrical and Electronics Engineering
Nowadays, healthy eating is very popular. People start to cook their own food from existing cooking ingredients. However, sometimes they do not know what food can be cooked from existing ingredients. Therefore, they cook the same food, resulting in monotonous eating and not enjoy cooking. This research article presents the design and development of a system for cooking recipe sharing and cooking recipe finding by an image of ingredients using deep learning techniques. Users can use the application on mobile devices to share cooking recipes. Moreover, users can take a picture of ingredients that users already have in the kitchen such as garlic, pork, vegetables, etc. and send that picture into the system to search for cooking recipes from existing ingredients. This process will make users convenient for searching cooking recipes. The main components of the system include (1) A mobile application for general users developed with React Native, which users can add cooking recipes and search for cooking recipes by entering the ingredient names. Also, the user can use the mobile device to take a picture of ingredients to find cooking recipes, (2) Web application developed on the MERN stack for system administrators, which system administrators can add keywords of the search term for ingredients and test the model that has been created, and (3) Deep convolutional neural network using the YOLO algorithm through the Darknet library for creating the image recognition model. The model has taught to be able to recognize 20 types of ingredients using 100 images of each type of ingredient. After training our model for 36,000 rounds, the model has an average loss of 0.0408 with the precision, recall and F1-score at 0.96, 0.98 and 0.97 respectively.
Date12/09/2019 120

บทความใกล้เคียง

Get alert for journal's news

Get alert for journal's news. You can recieve journal up-to-date information by giving your email to us.