[4Xin2] Poster session 2

[4Xin2-47] Deep learning toward new food texture analysis capable to discriminate foods having similar food textures, and to design food texture.

〇Ryusei Kaneko¹, Makoto Takemasa¹ (1.Tokyo Denki Univ.)

Keywords:Deep learning, Food texture analysis, Food texture design, Food 3D printer

Food texture is important for deliciousness, and dominating more than 60% of food taste. Deep learning was applied to (1) analyze and to (2) design food textures. (1) in the food texture analysis, Texture profile analysis (TPA) is generally used for food texture analysis based on compression test. TPA is an analysis method to extract several characteristic values based on the food compression test using the instrument, texture analyzer. It is quite difficult to analyze foods having similar food textures, compared with sensory test. By applying deep learning to food texture analysis based on the raw data of food compression test, we succeeded to distinguish and to analyze food textures even for the quite similar textures. (2) Food texture design has been performed based on empirical way, since the prediction is quite difficult due to the non-linearity of the stress-strain relationship of foods. Deep learning was again applied for this food texture design. The relation between the food internal structure and food texture, was learned using the food compression results of 3D printed food, some food structures could be predicted based on the learned data. This could be used as food texture design guideline.