Sharing Visual Features for Multiclass and Multiview Object Detection

We consider the problem of detecting a large number of different classes of objects in cluttered scenes. Traditional approaches require applying a battery of different classifiers to the image, at multiple locations and scales. This can be slow and can require a lot of training data since each classifier requires the computation of many different image features. In particular, for independently trained detectors, the (runtime) computational complexity and the (training-time) sample complexity scale linearly with the number of classes to be detected. We present a multitask learning procedure, based on boosted decision stumps, that reduces the computational and sample complexity by finding common features that can be shared across the classes (and/or views). The detectors for each class are trained jointly, rather than independently. For a given performance level, the total number of features required and, therefore, the runtime cost of the classifier, is observed to scale approximately logarithmically with the number of classes. The features selected by joint training are generic edge-like features, whereas the features chosen by training each class separately tend to be more object-specific. The generic features generalize better and considerably reduce the computational cost of multiclass object detection

1 Introduction

A long-standing goal of machine vision has been to build a system which is able to recognize many different kinds of objects in a cluttered world. Although the general problem remains unsolved, progress has been made on restricted versions of this goal. One succesful special case considers the problem of detecting individual instances of highly textured objects, such as magazine covers or toys, despite clutter, occlusion, and affine transformations. The method exploits features which are invariant to various transformations, yet which are very specific to a particular object [24], [31]. This can be used to solve tasks such as “find an object that looks just like this one,” where the user presents a specific instance, but it cannot be used to solve tasks such as “find an object that looks like a car,” which requires learning an appearance model of a generic car.