SO-DETR: Leveraging Dual-Domain Features and Knowledge Distillation for Small Object Detection

2025-04-11

AI Summary

Existing Transformer-based detectors suffer from limited small-object detection performance in UAV and remote sensing imagery due to (i) encoder architectures incapable of effectively fusing low-level detail features, and (ii) query initialization lacking explicit small-object awareness. To address these issues, we propose SO-DETR—a novel end-to-end detector. Its key contributions are: (1) a spatial-frequency dual-domain hybrid encoder enabling efficient cross-domain multi-scale feature fusion; (2) a dynamic anchor-guided query selection mechanism based on extended IoU, enhancing localization priors for small objects; and (3) a lightweight ResNet backbone coupled with hierarchical knowledge distillation for balanced accuracy-efficiency trade-offs. On VisDrone-2019-DET and UAVVaste, SO-DETR achieves state-of-the-art results with comparable computational cost—improving mAP by 3.2–4.7 percentage points over prior methods—and runs at 28 FPS on a Tesla V100 GPU.

Technology Category

• Computer Vision: Object Detection & Categorization
• Search and Optimization: Learning to Search
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Detection Transformer-based methods have achieved significant advancements in general object detection. However, challenges remain in effectively detecting small objects. One key difficulty is that existing encoders struggle to efficiently fuse low-level features. Additionally, the query selection strategies are not effectively tailored for small objects. To address these challenges, this paper proposes an efficient model, Small Object Detection Transformer (SO-DETR). The model comprises three key components: a dual-domain hybrid encoder, an enhanced query selection mechanism, and a knowledge distillation strategy. The dual-domain hybrid encoder integrates spatial and frequency domains to fuse multi-scale features effectively. This approach enhances the representation of high-resolution features while maintaining relatively low computational overhead. The enhanced query selection mechanism optimizes query initialization by dynamically selecting high-scoring anchor boxes using expanded IoU, thereby improving the allocation of query resources. Furthermore, by incorporating a lightweight backbone network and implementing a knowledge distillation strategy, we develop an efficient detector for small objects. Experimental results on the VisDrone-2019-DET and UAVVaste datasets demonstrate that SO-DETR outperforms existing methods with similar computational demands. The project page is available at https://github.com/ValiantDiligent/SO_DETR.

Problem

• Improving small object detection with dual-domain feature fusion
• Optimizing query selection for small objects using expanded IoU
• Reducing computational overhead via lightweight backbone and knowledge distillation

Innovation

• Dual-domain hybrid encoder integrates spatial and frequency domains
• Enhanced query selection uses expanded IoU for anchors
• Lightweight backbone with knowledge distillation boosts efficiency