Contrast R-CNN for Continual Learning in Object Detection

TL;DR

Contrast R-CNN introduces a novel continual learning approach for object detection that balances retaining old knowledge and learning new, using Proposal Contrast to improve robustness, validated on PASCAL VOC.

Contribution

We propose Contrast R-CNN, a new continual learning method for object detection that incorporates Proposal Contrast to better distinguish old and new instances.

Findings

Effective in retaining old knowledge while learning new objects

Outperforms existing methods on PASCAL VOC dataset

Robustness improved through Proposal Contrast mechanism

Abstract

The continual learning problem has been widely studied in image classification, while rare work has been explored in object detection. Some recent works apply knowledge distillation to constrain the model to retain old knowledge, but this rigid constraint is detrimental for learning new knowledge. In our paper, we propose a new scheme for continual learning of object detection, namely Contrast R-CNN, an approach strikes a balance between retaining the old knowledge and learning the new knowledge. Furthermore, we design a Proposal Contrast to eliminate the ambiguity between old and new instance to make the continual learning more robust. Extensive evaluation on the PASCAL VOC dataset demonstrates the effectiveness of our approach.