1 / 8

Automated detection of faces in images

Automated detection of faces in images. Sameer Jain Mehdi Mohseni Joy Rajiv. Flow Chart. Neural network. Skin color Based filtering. FLD. &. Regions in doubt. Low threshold Some non faces. +. Template matching. -. faces. +. High threshold All faces. Color Based Filtering.

mada
Download Presentation

Automated detection of faces in images

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. Automated detection of faces in images Sameer Jain Mehdi Mohseni Joy Rajiv

  2. Flow Chart Neural network Skin color Based filtering FLD & Regions in doubt Low threshold Some non faces + Template matching - faces + High threshold All faces

  3. Color Based Filtering V S H C. Garcia et al, IEEE Trans. Multim. 1999 • Hue Saturation Intensity space • Trained for face colors (8 vector) • Trained for non-face colors (32 vectors)

  4. Template Matching • Template of 6 sizes created (needed only smallest and 5th largest) • Segmented image into 2 halfs: • Top half smaller template • Bottom half larger template • Combined result • High threshold: miss some faces • Low threshold: get false hits

  5. Fisher Linear Discriminant • Procedure described in class • Worked on the template matching result with a lower threshold • Still see some false hits • Limitation due to: • Small training set • Face and non face not linearly separable

  6. w1 f1 w0 w2 f2 + +1:face +1 -1 -1: non face en=dn-wnHfn wn2 fn2 wn+1=wn-uenfn Neural network • Trained the system: • -1 for a non face • +1 for a face • Again saw some false hits • Limited success due to small training set

  7. Final Result of Routine • Combined results of FLD and neural network • Detect all faces in 4/7 images (~96% accuracy) • Approx time (~20s)

  8. Conclusion • Used • Color information • Shape information (template matching) • FLD • neural network approach • Use fact that the image set is limited • Fast algorithm (~20s) • Accurate algorithm (96% accuracy) on given test images

More Related