Learning from Noisy Anchors for One-stage Object Detection

TL;DR

This paper introduces a method to improve one-stage object detection by dynamically assessing anchor quality through cleanliness scores, reducing label noise and enhancing accuracy without extra computation.

Contribution

It proposes a novel approach to mitigate label noise in anchor-based detection by using dynamically estimated cleanliness scores for supervision and re-weighting.

Findings

Steady ~2% improvement on RetinaNet with various backbones.

Effectively reduces noise from imperfect anchor labels.

Enhances both localization and classification accuracy.

Abstract

State-of-the-art object detectors rely on regressing and classifying an extensive list of possible anchors, which are divided into positive and negative samples based on their intersection-over-union (IoU) with corresponding groundtruth objects. Such a harsh split conditioned on IoU results in binary labels that are potentially noisy and challenging for training. In this paper, we propose to mitigate noise incurred by imperfect label assignment such that the contributions of anchors are dynamically determined by a carefully constructed cleanliness score associated with each anchor. Exploring outputs from both regression and classification branches, the cleanliness scores, estimated without incurring any additional computational overhead, are used not only as soft labels to supervise the training of the classification branch but also sample re-weighting factors for improved localization and classification accuracy. We conduct extensive experiments on COCO, and demonstrate, among other things, the proposed approach steadily improves RetinaNet by ~2% with various backbones.