Wearable cameras have become more popular in recent years for capturing unscripted moments in the first-person, which help in analysis of the user's lifestyle. In this work, we aim to identify the daily food patterns of a person through recognition of places relating to food in person-focused images ("selfies"). This has the potential for a system that can assist with improvements to eating habits and prevention of diet-related conditions. In this paper, we use Siamese Neural Networks (SNN) to learn similarities between images with one-shot "food places" classification. We tested our proposed method with "MiniEgoFoodPlaces", using 15 food-related locations. The proposed SNN model with MobileNet achieved an overall classification accuracy of 76.74% and 77.53% on the validation and test sets of the "MiniEgoFoodPlaces" dataset, outperforming the base models such as ResNet50, InceptionV3 and InceptionResNetV2.
SARKER, M.M.K., BANU, S.F., RASHWAN, H.A., ABDEL-NASSER, M., SINGH, V.K., CHAMBON, S., RADEVA, P. and PUIG, D. 2019. Food places classification in egocentric images using Siamese neural networks. In Sabater-Mir, J., Torra, V., Aguiló, I. and González-Hidalgo, M. (eds.) Artificial intelligence research and development: proceedings of the 22nd International conference of the Catalan Association for Artificial Intelligence (CCIA 2019), 23-25 October 2019, Colònia de Sant Jordi, Spain. Frontiers in artificial intelligence and applications, 319. Amsterdam: IOS Press [online], pages 145-151. Available from: https://doi.org/10.3233/FAIA190117