Abstract:
Object detection is a fundamental problem in computer vision. Detection of faces, in
particular, is a critical part of face recognition and critical for systems which interact with
users visually.
We present a view-based approach implemented with artificial neural networks for face
detection. A retinally connected neural network examines small windows of an image, and
decides whether each window contains a face. The system arbitrates between multiple
networks to improve performance over a single network. We present a straightforward
procedure for aligning positive face examples for training. To collect negative examples,
we use a bootstrap algorithm, which adds false detections into the training set as training
progresses. This eliminates the difficult task of manually selecting nonface training
examples, which must be chosen to span the entire space of nonface images. Simple
heuristics, such as using the fact that faces rarely overlap in images, can further improve theaccuracy.
In the end we performed sensitivity analysis on the networks, and present empirical results
on a large test set.
