Aerial object detection improves by combining global and local features
LiG-DETR: Local-in-Global Reassembly in Latent Space for Aerial Object Detection
Computer Vision and Pattern Recognition
Summary
Detecting objects in aerial images is tricky because objects can be very small or very large, and pictures are often compressed, losing details. The authors created LiG-DETR, a method that uses both whole-picture features and zoomed-in local features to better spot objects of all sizes. Their system focuses on important areas and adjusts processing to save time, while improving accuracy for small and medium objects without hurting large object detection. This approach helps the detector keep detail lost during compression and combines information in one smooth process.
What this means in practice
- •For drone monitoring teams: Improve detection of small and medium objects in drone images by combining detailed local features with global scene understanding.
- •For satellite image analysts: Increase accuracy in automated object detection across various scales in satellite imagery by using a unified global-local feature framework.
Authors
Yupeng Zhang, Fangzhuo Gao, Juntao Cheng, Ziyi Zhao, Liang Wan, Ruize Han
Abstract
Aerial object detection faces substantial scale and density variations. Small objects are easily degraded by downsampling and feature compression, while medium and large objects require sufficient global context. Existing methods mainly follow two paradigms: image slicing provides clearer local evidence but relies on independent crop-level prediction and post-processing, whereas feature- and query-level optimization preserves unified inference but operates on already compressed full-image representations, limiting recovery of fine-grained information. This raises a key question: can aerial detection directly acquire high-fidelity local evidence before feature degradation and integrate it into a unified end-to-end framework? To this end, we propose LiG-DETR, an Efficient Global-Local Reassembly framework that reformulates image slicing as high-fidelity local feature acquisition. A shared encoder extracts global and locally magnified features, which are projected into the detector feature space. The projected local features are reassembled according to their original spatial locations to form a globally aligned local feature level, and a single DETR decoder jointly decodes global and local features. To reduce redundant computation, Context-Preserved Selective Reassembly focuses high-resolution encoding on informative regions while preserving a dense feature layout, and Density-Aware Adaptive Query Allocation adapts the decoder query budget using encoder proposal scores. Experiments show substantial gains on small and medium objects while retaining strong large-object performance, with favorable accuracy--efficiency trade-offs and improved cross-domain generalization. The code will be released.