大量の画像付き料理レシピからの学習による料理に含まれる素材推定

Bibliographic Information

Other Title
  • タイリョウ ノ ガゾウ ツキ リョウリ レシピ カラ ノ ガクシュウ ニ ヨル リョウリ ニ フクマレル ソザイ スイテイ
  • Ingredients Estimation in a Dish by Learning from a Large Number of Recipes with Images

Search this article

Abstract

In this report, we propose a method for estimating the ingredients contained in the dish in an input food image. Recently, the number of services that analyzes the preferences of an individual by recording daily meals, for improving of health, is increasing. Conventional works that try to realize such services, estimate the calorie from the name of the dish detected from the image. However, in practice, different ingredients may be used in the same kind of dishes, so only recognizing the type of a dish is insufficient. Therefore, we propose a method for estimating the ingredients in the dish in an input food image. Since the color and the shape features of an ingredient tend to appear in a dish, the proposed method extracts image feature from the food image and builds classifiers. As a result of an evaluation experiments, we confirmed that the performance of the classifier exceeds random classification.

IEICE Technical Report;IMQ2012-55,IEICE Technical Report;IE2012-159,IEICE Technical Report;MVE2012-116,IEICE Technical Report;WIT2012-65

Journal

References(1)*help

See more

Details

Report a problem

Back to top