nVFNet-RDC: Replay and Non-Local Distillation Collaboration for Continual Object Detection

TL;DR

This paper introduces nVFNet-RDC, a continual object detection method using teacher-student models with replay and feature distillation, achieving top results in recent challenge tracks.

Contribution

It presents a novel continual detection approach combining replay and non-local distillation within a teacher-student framework, setting new state-of-the-art performance.

Findings

Achieved 55.94% average mAP on CLVision Challenge Track 2.

Achieved 54.65% average mAP on CLVision Challenge Track 3.

First place solutions in recent continual object detection challenges.

Abstract

Continual Learning (CL) focuses on developing algorithms with the ability to adapt to new environments and learn new skills. This very challenging task has generated a lot of interest in recent years, with new solutions appearing rapidly. In this paper, we propose a nVFNet-RDC approach for continual object detection. Our nVFNet-RDC consists of teacher-student models, and adopts replay and feature distillation strategies. As the 1st place solutions, we achieve 55.94% and 54.65% average mAP on the 3rd CLVision Challenge Track 2 and Track 3, respectively.