Transformer method boosts tiny object detection accuracy on images
Reviving the Discarded High-Resolution Feature for Transformer-Based Tiny Object Detection
Computer Vision and Pattern Recognition
Summary
Detecting tiny objects in images is hard because they take up very few pixels and lose details when the image is simplified. The authors found a way to use the highest-detail features that are usually discarded because they are expensive to process. Their new method smartly combines these detailed features with other information, helping the detection system better recognize and locate tiny objects. This leads to better results on several benchmarks, especially for very small objects.
What this means in practice
- •For autonomous vehicle teams: Improve detection of small objects like pedestrians and cyclists in vehicle camera data by integrating detailed image features sensibly with transformers.
- •For drone surveillance operators: Enhance tiny object recognition in aerial images for monitoring and tracking using transformer models that revive and utilize high-resolution image details.
Authors
Zehao Wang, Si Wu, Guiping Cao, Xin Li, Yaowei Wang
Abstract
Tiny object detection remains challenging because tiny instances occupy only a few pixels, and their fine-grained details are easily lost during feature downsampling. Although the highest-resolution feature preserves rich local details for tiny objects, DETR-like detectors often discard it due to its large computational cost, while directly reintroducing this feature also brings dense background noise. To address this problem, we propose ReDHF, a novel framework that Revives the Discarded High-resolution Feature for tiny object detection. ReDHF has three cooperative components. Cross-Scale Deformable Fusion (CSDF) first constructs a semantically enhanced highest-resolution feature and injects its fine-grained details into the shallow feature under a moderate cost. Ellipse-Guided Shallow Supervision (EGSS) then guides the enhanced feature to learn more reliable classification and localization priors by applying geometry-adaptive auxiliary supervision. Full-Scale Query Interaction (FSQI) further enables decoder queries to access both low-level details and high-level semantic information, allowing them to aggregate richer cues for tiny objects. Extensive experiments show that ReDHF consistently improves multiple baselines across different DETR variants and achieves state-of-the-art performance on the evaluated benchmarks. The best ReDHF variants surpass their baselines by +1.4 AP on AI-TOD-v2, +1.6 AP on VisDrone, and +1.5 AP on SODA-D, with especially notable gains on tiny and small objects.