EW-DETR: Evolving World Object Detection via Incremental Low-Rank DEtection TRansformer

TL;DR

EW-DETR is a novel framework that enhances object detection in evolving environments by integrating incremental learning, domain adaptation, and unknown object detection without prior data, using DETR-based models.

Contribution

The paper introduces EW-DETR, a comprehensive framework with three modules for exemplar-free incremental learning, objectness decoupling, and unknown detection, applicable to DETR-based detectors.

Findings

EW-DETR outperforms existing methods on Pascal Series and Diverse Weather benchmarks.

It improves the FOGS score by 57.24%, demonstrating superior holistic performance.

The framework generalizes across DETR-based detectors, enabling effective operation in evolving-world settings.

Abstract

Real-world object detection must operate in evolving environments where new classes emerge, domains shift, and unseen objects must be identified as "unknown": all without accessing prior data. We introduce Evolving World Object Detection (EWOD), a paradigm coupling incremental learning, domain adaptation, and unknown detection under exemplar-free constraints. To tackle EWOD, we propose EW-DETR framework that augments DETR-based detectors with three synergistic modules: Incremental LoRA Adapters for exemplar-free incremental learning under evolving domains; a Query-Norm Objectness Adapter that decouples objectness-aware features from DETR decoder queries; and Entropy-Aware Unknown Mixing for calibrated unknown detection. This framework generalises across DETR-based detectors, enabling state-of-the-art RF-DETR to operate effectively in evolving-world settings. We also introduce FOGS (Forgetting, Openness, Generalisation Score) to holistically evaluate performance across these dimensions. Extensive experiments on Pascal Series and Diverse Weather benchmarks show EW-DETR outperforms other methods, improving FOGS by 57.24%.