IOR: Inversed Objects Replay for Incremental Object Detection

2024-06-07

AI Summary

To address catastrophic forgetting in incremental object detection caused by the absence of old-class samples, this paper proposes an inversion-based old-class sample replay mechanism that eliminates the need for auxiliary generative models. Our method leverages the detector’s own gradients to invert and synthesize high-fidelity old-class samples—bypassing the computational and storage overhead of training and maintaining separate generative models. We further introduce an enhanced replay strategy to improve sample reuse efficiency and incorporate high-value knowledge distillation, specifically transferring localization knowledge for challenging scenarios such as background occlusion. Evaluated on the MS COCO 2017 benchmark, our approach significantly improves detection performance on old classes while maintaining inference efficiency and accuracy. Crucially, it completely avoids redundant sample generation, achieving effective continual learning without external synthesis modules.

Technology Category

• Machine Learning: Life-Long and Continual Learning
• Computer Vision: Object Detection & Categorization
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Existing Incremental Object Detection (IOD) methods partially alleviate catastrophic forgetting when incrementally detecting new objects in real-world scenarios. However, many of these methods rely on the assumption that unlabeled old-class objects may co-occur with labeled new-class objects in the incremental data. When unlabeled old-class objects are absent, the performance of existing methods tends to degrade. The absence can be mitigated by generating old-class samples, but it incurs high costs. This paper argues that previous generation-based IOD suffers from redundancy, both in the use of generative models, which require additional training and storage, and in the overproduction of generated samples, many of which do not contribute significantly to performance improvements. To eliminate the redundancy, we propose Inversed Objects Replay (IOR). Specifically, we generate old-class samples by inversing the original detectors, thus eliminating the necessity of training and storing additional generative models. We propose augmented replay to reuse the objects in generated samples, reducing redundant generations. Moreover, we propose high-value knowledge distillation focusing on the positions of old-class objects overwhelmed by the background, which transfers the knowledge to the incremental detector. Extensive experiments conducted on MS COCO 2017 demonstrate that our method can efficiently improve detection performance in IOD scenarios with the absence of old-class objects.

Problem

• Incremental Object Detection
• Performance Degradation
• Sample Redundancy

Innovation

• Inverse Object Replay (IOR)
• Enhanced Replay Technique
• High-value Knowledge Distillation