InceptFood: A CNN Based Classification Approach for Recognizing Food Images
share
Food image recognition is one of the most optimistic applications of visual object recognition in the field of computer vision. Convolutional neural networks are at the core of most state-of-the-art computer vision solutions for a broad range of tasks. In this paper, we propose a novel Deep Learning based approach for recognizing food images. Based on Inception V3 model of Tensorflow platform, we use the transfer learning technology to retrain the final layer of the renowned Inception V3 architecture developed by Google for our classification approach. Data augmentation methods based on geometric transformation were implemented to enhance the volume of training images. Our approach shows promising results with an overall accuracy of 91% approximately in correctly recognizing food images while preventing the overfitting problem.
READ FULL TEXT