Random Boxes Are Open-world Object Detectors

TL;DR

This paper introduces RandBox, a novel open-world object detection method using random region proposals, which improves detection of both known and unknown objects by leveraging randomness to prevent training bias and enhance proposal exploration.

Contribution

RandBox is a new architecture that employs random proposals during training, outperforming existing methods in open-world object detection benchmarks.

Findings

RandBox surpasses state-of-the-art in Pascal-VOC, MS-COCO, and LVIS.

Random proposals prevent training bias towards known objects.

Unbiased training encourages better exploration of proposals.

Abstract

We show that classifiers trained with random region proposals achieve state-of-the-art Open-world Object Detection (OWOD): they can not only maintain the accuracy of the known objects (w/ training labels), but also considerably improve the recall of unknown ones (w/o training labels). Specifically, we propose RandBox, a Fast R-CNN based architecture trained on random proposals at each training iteration, surpassing existing Faster R-CNN and Transformer based OWOD. Its effectiveness stems from the following two benefits introduced by randomness. First, as the randomization is independent of the distribution of the limited known objects, the random proposals become the instrumental variable that prevents the training from being confounded by the known objects. Second, the unbiased training encourages more proposal explorations by using our proposed matching score that does not penalize the random proposals whose prediction scores do not match the known objects. On two benchmarks: Pascal-VOC/MS-COCO and LVIS, RandBox significantly outperforms the previous state-of-the-art in all metrics. We also detail the ablations on randomization and loss designs. Codes are available at https://github.com/scuwyh2000/RandBox.