[2Win5] Poster session 2

[2Win5-68] Development of new generation digital methods for food texture design based on deep learning

〇Ryusei Kaneko¹, Makoto Takemasa¹ (1.Graduate School of Tokyo Denki University)

Keywords:deep learning, food texture, food structure

More than 60% of palatability is dominated by food texture. Food texture originates not only from material properties but also from food structure. There is no guideline for food texture design, and empirical methods involving 'trial and error' are commonly used, which is a big challenge in food design due to too many factors affecting textures. If the food structure to realize desired food texture can be predicted based on machine learning, it is expected to make food texture design much more efficient. Based on collected texture data in relation to structure, it is now possible to predict the food structure resulting in given food texture, particularly multi-layered food as a simplest system. Applying this method, it is expected to be now possible to predict other structure such as block structure with the addition of larger dataset.