RODEO: Replay for Online Object Detection

TL;DR

This paper introduces RODEO, a novel online replay method for incremental object detection that addresses catastrophic forgetting and achieves state-of-the-art results on PASCAL VOC 2007 and MS COCO datasets.

Contribution

It presents a new memory replay mechanism enabling online learning of object detection with continual class addition and limited memory, advancing incremental learning capabilities.

Findings

Achieves state-of-the-art results on PASCAL VOC 2007

Achieves state-of-the-art results on MS COCO

Effectively mitigates catastrophic forgetting in online detection

Abstract

Humans can incrementally learn to do new visual detection tasks, which is a huge challenge for today's computer vision systems. Incrementally trained deep learning models lack backwards transfer to previously seen classes and suffer from a phenomenon known as $"catastrophic forgetting."$ In this paper, we pioneer online streaming learning for object detection, where an agent must learn examples one at a time with severe memory and computational constraints. In object detection, a system must output all bounding boxes for an image with the correct label. Unlike earlier work, the system described in this paper can learn this task in an online manner with new classes being introduced over time. We achieve this capability by using a novel memory replay mechanism that efficiently replays entire scenes. We achieve state-of-the-art results on both the PASCAL VOC 2007 and MS COCO datasets.