1 / 15

Face-Detection using Maximal Rejection Classification and Color Techniques

Face-Detection using Maximal Rejection Classification and Color Techniques. Group 11 Sam Mazin & Priti Balchandani. Motivation. Michael Elad’s presentation on rejection-based techniques Felt a linear classifier (MRC) would be fastest and thus most practical in real-world applications.

sabin
Télécharger la présentation

Face-Detection using Maximal Rejection Classification and Color Techniques

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. Face-Detection using Maximal Rejection Classification and Color Techniques Group 11 Sam Mazin & Priti Balchandani

  2. Motivation • Michael Elad’s presentation on rejection-based techniques • Felt a linear classifier (MRC) would be fastest and thus most practical in real-world applications

  3. High-level Design Maximal Rejection Classifier (MRC) Color Rejection Morphological Processing FACES

  4. Maximal Rejection Background Training: Backgrounds d1 Faces (Convex) d2 θ

  5. Maximal Rejection Classifying: Background d1n ? MRC Projection 15x15 Test Block Last θ Face Not last θ: Project with θn+1 d2n Very fast θn

  6. Color Rejection • MRC was not the greatest on its own; many false positives remained • Tried RGB segmentation  not great • Y-Cb-Cr space showed promise • Decided to take mean Cb and Cr values of all 15x15 blocks (ignored Y to avoid intensity bias)

  7. Color Rejection using FLD

  8. Optimal Cb-Cr FLD Projection Projection Density: p(w) Faces Backgrounds w W*=max(FLD_proj(Faces))

  9. Morphological Processing Maximal Rejection Classifier (MRC) Color Rejection Morphological Processing FACES

  10. Morphological Processing • Combined 3 resolution levels into one • Dilated and performed centroid search to get rid of repeated face detections

  11. Finally…

  12. Desperate times… • Monday night, 10pm • Still had 5-6 false positives popping up • Decided to implement the “Look Down Method” • Reran tests: made good scores better, but bad scores worse

  13. Conclusion • MRC is fast but not 100% reliable (probably due to lack of data) • Color rejection helped significantly, Cb-Cr good means of classification • Morphological processing necessary for repeated detections • Spent too much time tweaking the MRC

More Related