UAVPairs: A Challenging Benchmark for Match Pair Retrieval of Large-scale UAV Images

TL;DR

This work tackles match-pair retrieval for large-scale UAV imagery and its impact on SfM-based 3D reconstruction. It introduces UAVPairs, a geometry-grounded benchmark built with SfM-derived annotations, and a training pipeline combining batched nontrivial sample mining with a ranked list loss to improve global discrimination of deep global features. Experiments across three UAV datasets show that models trained on UAVPairs with the ranked list loss achieve higher retrieval accuracy and substantially better view-graph connectivity and 3D reconstruction quality than baselines or models trained on existing datasets. The proposed benchmark and training pipeline offer a practical, scalable solution for robust UAV image matching in repetitive or weak-texture scenes, with public availability for community use.

Abstract

The primary contribution of this paper is a challenging benchmark dataset, UAVPairs, and a training pipeline designed for match pair retrieval of large-scale UAV images. First, the UAVPairs dataset, comprising 21,622 high-resolution images across 30 diverse scenes, is constructed; the 3D points and tracks generated by SfM-based 3D reconstruction are employed to define the geometric similarity of image pairs, ensuring genuinely matchable image pairs are used for training. Second, to solve the problem of expensive mining cost for global hard negative mining, a batched nontrivial sample mining strategy is proposed, leveraging the geometric similarity and multi-scene structure of the UAVPairs to generate training samples as to accelerate training. Third, recognizing the limitation of pair-based losses, the ranked list loss is designed to improve the discrimination of image retrieval models, which optimizes the global similarity structure constructed from the positive set and negative set. Finally, the effectiveness of the UAVPairs dataset and training pipeline is validated through comprehensive experiments on three distinct large-scale UAV datasets. The experiment results demonstrate that models trained with the UAVPairs dataset and the ranked list loss achieve significantly improved retrieval accuracy compared to models trained on existing datasets or with conventional losses. Furthermore, these improvements translate to enhanced view graph connectivity and higher quality of reconstructed 3D models. The models trained by the proposed approach perform more robustly compared with hand-crafted global features, particularly in challenging repetitively textured scenes and weakly textured scenes. For match pair retrieval of large-scale UAV images, the trained image retrieval models offer an effective solution. The dataset would be made publicly available at https://github.com/json87/UAVPairs.

Figures (13)

• Figure 1: UAV images of various scenes
• Figure 2: UAV imagThe reconstructed 3D models of various scenes
• Figure 3: The overview workflow of the training pipeline.
• Figure 4: A training batch example generated based on the batched nontrivial sample mining strategy
• Figure 5: Three possibilities for the negative of a triplet.