Mitigate the classification ambiguity via localization-classification sequence in object detection

In anchor-based detectors, the confidence scores and label-assignment results for the classification task are determined by the unrefined anchors rather than the final-refined boxes, which causes classifica-tion ambiguity due to the lack of correlation between the classification and localization tasks. In this paper, we investigate the classification ambiguity thoroughly via extensive experiments, and present the localization-classification sequence detector (LCSDet) that performs localization and classification in or-der, bridging the gap between them. To achieve this, the refinement-aware (RA) classification branch and RA assignment are proposed in LCSDet. In inference, the RA classification branch rectifies the fea-ture misalignment and directly classifies the refined anchors. During training, the RA assignment tackles the training instability, narrows the location-quality gap and assigns the refined anchors to ground-truth objects. Comprehensive experiments indicate that the LCSDet can effectively mitigate the classification ambiguity and achieve stable improvement across different baselines.(c) 2023 Elsevier Ltd. All rights reserved.

Automated summary

In anchor-based detectors, classification ambiguity caused by the lack of correlation between classification and localization tasks is investigated. A localization-classification sequence detector (LCSDet) is proposed to bridge this gap, with refinement-aware classification branch and assignment. The LCSDet effectively mitigates the classification ambiguity and achieves stable improvement across different baselines.