A new 21.6K-image UAV benchmark with SfM-based pair labels plus a batched nontrivial mining strategy and ranked list loss improves match pair retrieval and downstream 3D reconstruction.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
UAVPairs: A Challenging Benchmark for Match Pair Retrieval of Large-scale UAV Images
A new 21.6K-image UAV benchmark with SfM-based pair labels plus a batched nontrivial mining strategy and ranked list loss improves match pair retrieval and downstream 3D reconstruction.