Automatic Construction of Feature Extraction Process Based on Combinatorial Optimization of Image Processing Filters

2017 Fiscal Year Final Research Report

• Project/Area Number: 15K16029
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Perceptual information processing
• Research Institution: Yokohama National University (2016-2017); University of Tsukuba (2015)
• Principal Investigator: Shirakawa Shinichi 横浜国立大学, 大学院環境情報研究院, 講師 (90633272)
• Project Period (FY): 2015-04-01 – 2018-03-31
• Keywords: 画像認識 / 画像特徴量 / 機械学習 / 進化計算

Outline of Final Research Achievements

In this research, we developed a novel technique that constructs efficient image feature extraction process by using a black-box optimization method based on combinatorial optimization of basic image processing units such as image processing filters. Further, we extended and applied the idea of this technique to convolutional neural networks (CNN) that show high performance on image recognition tasks. From the numerical experiments of image recognition, we confirmed that the proposed methods outperform the existing CNN based methods.

Free Research Field

知能情報処理，人工知能，ソフトコンピューティング