Project 2: Object detection by Antiface projections

 

 

Description

•         Implement the Antiface method for detection.

•         Implement the following extensions of Antifaces:

–       In the original method an image is classified as object if it is accepted by all antiface detectors, and the detection process is sequential. Change the detection process so that each antiface detector is applied to all sub-images. Implement the new acceptance rule: a sub-image is classified as object if it is accepted by 80% of antiface detectors.

–       Project the training images of the class on 10D antiface space. Find the parameters of Gaussian distribution that best describe the projections (using Maximum Likelihood Estimation). Since detectors are independent of each other, the covariance matrix can be assumed diagonal. Perform recognition using Naïve Bayes.

–       Project each image onto 20D antiface space, spanned by antiface detectors and train SVM in this space. Recognition should be done in similar way (projection onto antiface space and SVM classification).

Results

The detection results should be presented in the form of ROC curve, which shows the performance of the detection method by changing the threshold that controls the detection score. The x-axis is the number of true positives (faces found by the system) divided by the total number of faces in images (from the ground truth). The y- axis is (1- number of false positive, divided by the total number of non-face sub-images). False positive is a sub-image that the system detected as a face but it’s not marked as a face in the ground truth file.

 

Data sets

The system should be tested on  CBCL face set. This data set contains images for training and for testing.

 

SVM Package

Matlab version

C version

 

Useful Links

Class lecture on object detection

Slides explaining the antiface method

SVM tutorial

SVM slides

Matlab statistics toolbox.  I suggest implementing this project in Matlab using its statistics toolbox. It contains functions for MLE, Likelihood, and other useful stuff.

 

Status

Claimed by Shira and Ariel Gorfinkel