Survey of Different Approaches Used for Food Recognition

Abstract

Food recognition is an ever-growing field gaining rapid momentum in the past couple of years. Various approaches have been implemented to get accurate results by correctly identifying the food item. Traditional methods like the implementation of neural networks, SVMs, HMMs utilizing hand-crafted features of the large data-sets of food images are one way of developing food recognition systems. To improve the accuracy, modern methods using newer concepts of convolutional neural networks and deep learning which avoid the use of hand-crafted features are being implemented to build even better food recognition systems. These newer methods require huge data-sets of images of food items to work with to obtain good results. Besides approaches based on image recognition, other innovative images are also being explored for recognizing food images. Food items are being recognized using the cutting sounds, acoustic sensors, electronic tongues and so on.