LocNet: Improving Localization Accuracy for Object Detection

TL;DR

LocNet introduces a probabilistic approach using CNNs to improve object localization accuracy, significantly enhancing detection performance at high IoU thresholds and working independently of box proposal methods.

Contribution

It presents a novel CNN-based localization method that assigns conditional probabilities to region boundaries, boosting high-precision detection accuracy.

Findings

Significant improvement in mAP at high IoU thresholds on PASCAL VOC2007.

Easily integrates with existing detection systems to enhance performance.

Achieves high detection accuracy using sliding windows, independent of box proposals.

Abstract

We propose a novel object localization methodology with the purpose of boosting the localization accuracy of state-of-the-art object detection systems. Our model, given a search region, aims at returning the bounding box of an object of interest inside this region. To accomplish its goal, it relies on assigning conditional probabilities to each row and column of this region, where these probabilities provide useful information regarding the location of the boundaries of the object inside the search region and allow the accurate inference of the object bounding box under a simple probabilistic framework. For implementing our localization model, we make use of a convolutional neural network architecture that is properly adapted for this task, called LocNet. We show experimentally that LocNet achieves a very significant improvement on the mAP for high IoU thresholds on PASCAL VOC2007 test set and that it can be very easily coupled with recent state-of-the-art object detection systems, helping them to boost their performance. Finally, we demonstrate that our detection approach can achieve high detection accuracy even when it is given as input a set of sliding windows, thus proving that it is independent of box proposal methods.