1 / 16

Finding Bucky

Finding Bucky. Identifying unapproved logo representations. James Buchen , David Mateo, Liang Zheng Gooi. Motivation. D etect altered versions of Bucky logo Maintain UW - Madison Athletics brand Develop method to be applied to any brand or logo. Guidelines - Approved.

merry
Télécharger la présentation

Finding Bucky

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Finding Bucky Identifying unapproved logo representations James Buchen, David Mateo, Liang Zheng Gooi

  2. Motivation • Detect altered versions of Bucky logo • Maintain UW - Madison Athletics brand • Develop method to be applied to any brand or logo

  3. Guidelines - Approved

  4. Guidelines - Not approved

  5. Two Step Process • Use the Viola-Jones image detector • Weak Classifier model • Requires Image training • Use SIFT, SURF, or FAST • Flagged images will be compared directly • Approve the image based on number of feature points

  6. Viola-Jones Image Detector • Complex classifier • Uses cascading weak classifiers • Trained using positive and negative images • Feature values are sum of pixels in clear rectangles subtracted from sum of pixels within shaded rectangles Feature Types

  7. Method • Input official Bucky logo and website screenshot • Generate Bucky variations • Collect other variations from internet • Train viola-jones image detector • Identify potential Bucky objects • Compare objects and official Bucky using SIFT • Generate output image

  8. Positive Images:GeneratedCollected

  9. Other technologies we looked into • Extract cartoon from other image • Create occlusion image • SURF detector, Edge Feature (EF) • “Content Based Image Retrieval through Object Extraction and Querying”

  10. Results-Poor

  11. Results-Good

  12. Sift-Negative • Altered Aspect Ratio produces negative response • Obscured Bucky only has ~half the feature points matching

  13. Sift-Positive • Almost all feature points match • Invariant to scaling

  14. Future Improvements • Increase variety of generated images • Increase number of collected images • Tune the detector • Choose decision threshold for SIFT • Develop automatic website crawler to scan many websites for Unapproved Bucky images.

  15. Thank you! Questions?

  16. References [1] Viola, P., and M. J. Jones, "Rapid Object Detection using a Boosted Cascade of Simple Features". Proceedings of the 2001 IEEE Computer Society Conference. Volume 1, 15 April 2001, pp. I-511–I-518. [2] Ojala, T., M. Pietikainen, and T. Maenpaa, "Multiresolution Gray-scale and Rotation Invariant Texture Classification With Local Binary Patterns". IEEE Transactions on Pattern Analysis and Machine Intelligence. Volume 24, No. 7 July 2002, pp. 971–987. [3] Dalal, N., and B. Triggs, "Histograms of Oriented Gradients for Human Detection". IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Volume 1, (2005), pp. 886–893.

More Related