Towards a category-extended object detector with limited data. (December 2022)

• Record Type: Journal Article
• Authors: Zhao, Bowen; Chen, Chen; Xiao, Xi; Xia, Shutao
• Is Part Of: Pattern recognition. Volume 132(2022)
• Publication Date: 2022-12
• Subjects: Object detector -- Category-extended -- Limited data -- Multi-dataset; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2022.108943
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: Focusing on learning category-extended detector with limited data. Meeting the needs of real-world applications. Conflict-free loss leads to acceptable unified detectors in one training round. A well designed retraining phase further improve performance. Extensive experiments demonstrate the effectiveness of our method. Abstract: Object detectors are typically learned on fully-annotated training data with fixed predefined categories. However, categories are often required to be increased progressively. Usually, only the original training set annotated with old classes and some new training data labeled with new classes are available in such scenarios. Based on the limited datasets, a unified detector that can handle all categories is strongly needed. We propose a practical scheme to achieve it in this work. A conflict-free loss is designed to avoid label ambiguity, leading to an acceptable detector in one training round. To further improve performance, we propose a retraining phase in which Monte Carlo Dropout is employed to calculate the localization confidence to mine more accurate bounding boxes, and an overlap-weighted method is proposed for making better use of pseudo annotations during retraining. Extensive experiments demonstrate the effectiveness of our method.