Towards a category-extended object detector with limited data

TL;DR

This paper introduces a practical method for training a unified object detector capable of recognizing both old and new categories with limited data, using a conflict-free loss and confidence-based retraining.

Contribution

It proposes a novel scheme combining a conflict-free loss and Monte Carlo Dropout-based retraining to extend object detectors to new categories with limited data.

Findings

Effective in handling incremental category learning

Improves detection accuracy with limited data

Demonstrates robustness across experiments

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.