Version 22 (Anthony Rowe, 08/23/2013 03:09 am)

1 17 Anthony Rowe
h2. Viola Jones Face Detector
2 17 Anthony Rowe
3 17 Anthony Rowe
4 21 Anthony Rowe
The Viola-Jones sample project that comes with the CMUcam3 is an example of a lightweight face detector. The algorithm is based on the well known paper “Robust Real-Time Face Detection” by P. Viola and M. Jones from 2004 attachment:viola-ijcv04.pdf. The implementation on the CMUcam3 will return coordinates for boxes where it detects a face in the image.  If you test this code with a relatively uniform background (like a white wall), it works reasonably well.  The images are trained from the CMU face database such that they generalize to all faces. Sometimes certain faces with features not found in the database can confuse the detector.
5 1 Anthony Rowe
6 1 Anthony Rowe
!face.jpg! !face1.jpg!
7 1 Anthony Rowe
8 22 Anthony Rowe
attachment:viola-ijcv04.pdf introduces a novel technique to detect faces in real-time and with very high detection rate. It is essentially a feature-based approach in which a classifier is trained for Haar-like rectangular "features": selected by "Adaboost": The test image is scanned at different scales and positions using a rectangular window, and the regions which pass the classifier are declared as faces. One of the major contributions of this paper is the extremely rapid computation of these features using the concept of Integral Image, which enables the detection in real-time. Additionally, instead of learning a single classifier and computing all the features for all the scanning windows in the image, a number of classifiers are learnt which are put together in a series to form a cascade. The classifiers in the beginning of the cascade are simpler and consist of smaller numbers of features. However, as one proceeds in the cascade, the classifiers become more complex. A region is reported as detection only if it passes all the classifier stages in the cascade. If it is
9 21 Anthony Rowe
rejected at any stage, it is discarded and not processed further. This way, the easier patches in the image which the “cascade of classifiers” is sure of not being a face, are rejected very early while the difficult regions are operated on by more complex classifiers. This greatly speeds up the detection process without compromising on the accuracy and provides high detection rate. This overall system provides performance comparable to the existing best face detector systems (Rowley et al., 1998 attachment:CVPR00.pdf; Schneiderman and Kanade, 2000 attachment:nips00.pdf; Roth at al., 2000 attachment:rowley-ieee.pdf) but with orders of magnitudes faster than any of these systems. On a conventional desktop, it can detect faces at 15 frames per second.
10 5 Anthony Rowe
11 6 Anthony Rowe
More information on the Viola-Jones face detector and its CMUcam3 implementation can be found in our "CC3 Face Detector": document.
12 5 Anthony Rowe
13 17 Anthony Rowe
*Extra Utilities and Images*
14 17 Anthony Rowe
* attachment:generate_feat_in_struct_for_C.m  
15 20 Anthony Rowe
** This Matlab file prints the learned adaboost model in a text file so that it can be easily imported in vj.h 
16 17 Anthony Rowe
* attachment:get_scaled_feature.m
17 17 Anthony Rowe
** This file used by generate_feat_in_struct_for_C.m
18 21 Anthony Rowe
* Test images we used for the viola-jones face detector
19 21 Anthony Rowe
** "Face DB": "Face DB2":