Visual Event Learning with Web Resources

2015 Fiscal Year Final Research Report

• Project/Area Number: 26540081
• Research Category: Grant-in-Aid for Challenging Exploratory Research
• Allocation Type: Multi-year Fund
• Research Field: Perceptual information processing
• Research Institution: Nagoya University
• Principal Investigator: KATO Jien 名古屋大学, 情報科学研究科, 准教授 (70251882)
• Project Period (FY): 2014-04-01 – 2016-03-31
• Keywords: イベント認識 / 行動認識 / 学習データ収集の省力化 / 適応学習 / 近似スパースコーディング / 画像・映像検索 / 特徴次元選択 / データ選択

Outline of Final Research Achievements

The objective of this research is to develop a low cost event learning framework to enable easy event learning and recognition. The proposed framework has the following three components: (1) web data collecting, which helps to prepare learning data efficiently; (2) robust event learning, which guarantees the event recognition performance; and (3) domain adaption, which helps to transform the event models learned from the web domain to the target domain. In our work, we developed (1) a flexible event retrieval approach by integrating image recognition and nature language processing; (2) an approximate sparse coding based high performance event recognition approach; and (3) a feature selection based domain transform approach for adapting event model between different domains. The proposed objectives of this research have been mostly achieved.

Free Research Field

画像・映像の内容理解