Research on High Performance 3D Shape Retrieval under Diversified Queries including 2D Inputs and Semantic Attributes such as Holes and Concavities

2016 Fiscal Year Final Research Report

• Project/Area Number: 26280038
• Research Category: Grant-in-Aid for Scientific Research (B)
• Allocation Type: Partial Multi-year Fund
• Section: 一般
• Research Field: Multimedia database
• Research Institution: Toyohashi University of Technology
• Principal Investigator: AONO MASAKI 豊橋技術科学大学, 工学(系)研究科(研究院), 教授 (00372540)
• Co-Investigator(Renkei-kenkyūsha): TATSUMA Atsushi 豊橋技術科学大学, 大学院工学研究科, 助教 (60711166)
• Project Period (FY): 2014-04-01 – 2017-03-31
• Keywords: 情報検索 / 機械学習 / 3D / 特徴量

Outline of Final Research Achievements

We have developed high performance 3D shape retrieval called MRLBP, which stands for Multi-Resolution Local Binary Pattern. MRLBP has brought us, to our knowledge, the best search performance which can run without prior training. We have submitted a patent on MRLBP. 3D shape retrieval usually requires an arbitrary 3D shape as the query, but this is sometimes very burdensome. To cope with this problem, we have developed a novel method for 3D shape retrieval from an arbitrary 2D sketch, which we call OPHOG (Overlapped Pyramidal Histogram of Oriented Gradients). With OPHOG, we achieved world number 1 record for SHREC2014 international contest. For partial 3D shape retrieval, we have developed a new method based on 3D point clouds with a new feature vector, which we submitted another patent.

Free Research Field

マルチメディア検索