Characterizing the performance of an image-based recognizer for planar mechanical linkages in textbook graphics and hand-drawn sketches
详细信息 查看全文
- 作者：Matthew Eicholtz meicholt@andrew.cmu.edu" class="auth_mail" title="E-mail the corresponding author ; Levent Burak Kara ; lkara@cmu.edu" class="auth_mail" title="E-mail the corresponding author
- 关键词：Computer vision ; Evolutionary multiobjective optimization ; Image processing ; Kinematic modeling ; Object recognition ; Sketch recognition
- 刊名：Computers & Graphics
- 出版年：2015
- 出版时间：November 2015
- 年：2015
- 卷：52
- 期：Complete
- 页码：1-17
- 全文大小：9209 K
文摘
- •
We systematically evaluate a framework for recognizing mechanical linkages in images.
- •
We introduce user effort ratio to assess overall benefit over manual model creation.
- •
New rigid body detection and graph–theoretic constraints enhance performance.
- •
Top-1 accuracy for textbook images is 70% with user effort of 10% on average.
- •
Performance on sketches is lower overall, but still reduces user effort by 72% on average.