Research map / Detection and segmentation / Open-set and few-shot detection

DON’T NEED RETRAINING: A Mixture of DETR and Vision Foundation Models for Cross-Domain Few-Shot Object Detection

NeurIPS 2025 · Detection and segmentation › Open-set and few-shot detection · approach: Domain-adaptive object detection

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes a training-free mixture of a well-trained DETR detector and vision foundation models for cross-domain few-shot object detection, combining DETR's localization with VFMs' generalization.

Problem

Cross-domain few-shot object detection (CD-FSOD) needs detectors that both generalize to unseen domains from few annotated target samples and localize accurately; well-trained detectors localize well but generalize poorly, while vision foundation models generalize better but localize less well.

Approach

Instead of retraining, combine a well-trained DETR detector with vision foundation models in a mixture so that each contributes its strength (localization from DETR, generalization from VFMs), using few target-domain samples without retraining.

Experimental setup

Builds on: DETR

Closest papers by meaning

• Weakly Supervised Few-Shot Object Detection with DETR (AAAI 2024 · similarity 0.86)
• Improving Single Domain-Generalized Object Detection: A Focus on Diversification and Alignment (CVPR 2024 · similarity 0.86)
• Fine-Grained Prototypes Distillation for Few-Shot Object Detection (AAAI 2024 · similarity 0.86)
• Few-Shot Object Detection with Foundation Models (CVPR 2024 · similarity 0.86)
• Domain-RAG: Retrieval-Guided Compositional Image Generation for Cross-Domain Few-Shot Object Detection (NeurIPS 2025 · similarity 0.86)
• Frozen-DETR: Enhancing DETR with Image Understanding from Frozen Foundation Models (NeurIPS 2024 · similarity 0.86)