Local Decorrelation For Improved Detection

TL;DR

This paper introduces an efficient local decorrelation transform for features that enhances the performance of orthogonal decision trees in object detection, significantly reducing false positives and outperforming oblique trees.

Contribution

The authors propose a novel local decorrelation feature transform that improves detection accuracy while maintaining computational efficiency compared to oblique decision trees.

Findings

Orthogonal trees with decorrelated features outperform oblique trees.

Nearly tenfold reduction in false positives on Caltech Pedestrian Dataset.

Significant accuracy improvements over previous state-of-the-art methods.

Abstract

Even with the advent of more sophisticated, data-hungry methods, boosted decision trees remain extraordinarily successful for fast rigid object detection, achieving top accuracy on numerous datasets. While effective, most boosted detectors use decision trees with orthogonal (single feature) splits, and the topology of the resulting decision boundary may not be well matched to the natural topology of the data. Given highly correlated data, decision trees with oblique (multiple feature) splits can be effective. Use of oblique splits, however, comes at considerable computational expense. Inspired by recent work on discriminative decorrelation of HOG features, we instead propose an efficient feature transform that removes correlations in local neighborhoods. The result is an overcomplete but locally decorrelated representation ideally suited for use with orthogonal decision trees. In fact, orthogonal trees with our locally decorrelated features outperform oblique trees trained over the original features at a fraction of the computational cost. The overall improvement in accuracy is dramatic: on the Caltech Pedestrian Dataset, we reduce false positives nearly tenfold over the previous state-of-the-art.