
MARQS: Media Album Retrieval by Query Sketch

In an effort to make human computer interaction handle more forms of human-human interaction, sketching has recently become an important means of interacting with computer systems. Current feature-based recognition systems use single stroke features which either constrain drawing style or are domain-specific. Other systems use geometric recognition techniques to define symbols hierarchically; however, some symbols are difficult to describe geometrically. We present here an alternative recognition technique that uses domain-independent and style-independent global features to recognize an entire sketch symbol. Furthermore, we have integrated two different recognition algorithms, creating an unobtrusive learning method that can recognize shapes after only a single example but can learn from query sketches, thus improving accuracy over time. As a proof-of-concept application, we created MARQS, a system that uses sketches to query existing media albums. The system yielded an average search rank of 1.51, indicating that the correct sketch is presented as either the top or second search result on average.
Authors
Brandon PaulsonTracy Hammond
Publications
- Paulson, B. and Hammond, T. MARQS: Retrieving Sketches Using Domain-and Style-Independent Features Learned from a Single Example Using a Dual-Classifier. Journal of Multi-Modal User Interfaces, 2008. [HTML]
- Paulson, B. and Hammond, T. MARQS: Retrieving Sketches Using Domain- and Style-Independent Features Learned from a Single Example Using a Dual-Classifier. In Proceedings of IWUMUI, 2007. [PDF] [PPT Presentation] [MPEG]
- Paulson, B. and Hammond, T. MARQS: Media album retrieval by query sketch. (video) AAAI-07 AI Video Competition, 2007. (nominated: Best Short Video) [MOV] [AVI]*
Poster Presentations
- Paulson, B. and Hammond, T. Sketch Recognition Using Global Features. Student Research Week, Texas A&M University. March 26-30, 2007. (2nd place session winner) [PDF]
- Paulson, B. and Hammond, T. Sketch Recognition Using Global Features. Industry Affiliates Program, Texas A&M University. March 7, 2007. [PDF]
* - videos may require the free TechSmith Screen Capture Codec