1 / 8

REU: Week 9

REU: Week 9. Overview. Goal: To create a more robust method of classification. Solution: (This Week) Generate Histograms Apply early fusion on histograms. Precomputing Histograms. Generate descriptor vectors. Cluster these descriptors. Generate ‘bag of words’ histograms.

amanda
Télécharger la présentation

REU: Week 9

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. REU: Week 9

  2. Overview • Goal: To create a more robust method of classification. • Solution: (This Week) • Generate Histograms • Apply early fusion on histograms.

  3. Precomputing Histograms • Generate descriptor vectors. • Cluster these descriptors. • Generate ‘bag of words’ histograms. • These histograms can be combined in a method called early fusion. • Helpful because it will cut down on run times for final run(s).

  4. Histograms • Harris-Laplace and Dense Sampling detectors • SIFT • Scale invariant (only uses intensity channel) • Opponent SIFT • Includes information other than intensity channel, the opponent channels (not intensity invariant) • rgSIFT • Chromacity components r and g describe color information while being intensity invariant.

  5. Fusion • Complementary descriptors chosen to be more robust. • Quite a few papers suggest this gives better results than a single detector. • These histograms are concatenated together before going into the SVM. • Concatenation can be as effective as averaging in circumstances such as these.

  6. Results • The last of the descriptors are being computed and will be available to run histograms on over the weekend. • Code is in place to generate the histograms, as well as ‘fuse’ them before going into the SVM. • Actual results should be available for next week.

  7. Next Week • Results of high-level feature ‘Doorway’ using fusion of histograms. • Combine cross validation code with this week’s work to get some truly meaningful results. • Meaningful in the sense that they should give an accurate portrayal of accuracies to expect.

  8. Questions?

More Related