Search papers, labs, and topics across Lattice.
The paper introduces Evolving World Object Detection (EWOD), a new paradigm addressing incremental learning, domain adaptation, and unknown object detection without access to prior data. To tackle EWOD, the authors propose EW-DETR, a framework that enhances DETR-based detectors with Incremental LoRA Adapters, a Query-Norm Objectness Adapter, and Entropy-Aware Unknown Mixing. Experiments on Pascal Series and Diverse Weather benchmarks demonstrate that EW-DETR significantly outperforms existing methods, achieving a 57.24% improvement in the proposed FOGS (Forgetting, Openness, Generalisation Score) metric.
Forget catastrophic forgetting: EW-DETR enables object detectors to learn new classes and adapt to new domains without prior data, boosting overall performance by 57%.
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%.
