Enhance Then Search: An Augmentation-Search Strategy with Foundation Models for Cross-Domain Few-Shot Object Detection

2025-04-06

AI Summary

Cross-domain few-shot object detection (CD-FSOD) suffers from weak cross-domain generalization of base models and heavy reliance on costly retraining. Method: We propose a lightweight “Augment–Search” two-stage adaptation framework: (1) multi-strategy image augmentation—integrating CutMix, Mosaic, and ColorJitter—to enrich source-domain representation; (2) a differentiable, grid-based subdomain search algorithm that efficiently locates the optimal adaptation subdomain within the parameter space of the grounding model (GroundingDINO), enabling fine-tuning-free dynamic domain transfer. Contribution/Results: This work pioneers the coupling of structured subdomain search with data augmentation, establishing a lightweight parameter-space navigation paradigm that eliminates dependence on retraining. Our method achieves an average +5.2% mAP improvement across multiple CD-FSOD benchmarks, supports zero-shot domain transfer, and significantly reduces deployment overhead. Code is publicly available.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Search and Optimization: Learning to Search
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Foundation models pretrained on extensive datasets, such as GroundingDINO and LAE-DINO, have performed remarkably in the cross-domain few-shot object detection (CD-FSOD) task. Through rigorous few-shot training, we found that the integration of image-based data augmentation techniques and grid-based sub-domain search strategy significantly enhances the performance of these foundation models. Building upon GroundingDINO, we employed several widely used image augmentation methods and established optimization objectives to effectively navigate the expansive domain space in search of optimal sub-domains. This approach facilitates efficient few-shot object detection and introduces an approach to solving the CD-FSOD problem by efficiently searching for the optimal parameter configuration from the foundation model. Our findings substantially advance the practical deployment of vision-language models in data-scarce environments, offering critical insights into optimizing their cross-domain generalization capabilities without labor-intensive retraining. Code is available at https://github.com/jaychempan/ETS.

Problem

• Enhancing cross-domain few-shot object detection performance
• Optimizing sub-domain search with augmentation strategies
• Improving vision-language models in data-scarce environments

Innovation

• Image-based data augmentation enhances foundation models
• Grid-based sub-domain search optimizes performance
• Efficient parameter configuration from foundation models