This is the pytorch implementation for the submitted paper "A Scene Graph Encoding and Matching Network for UAV Visual Localization".
The detailed implementation of the proposed network (as the following figures show) can be find in .\sceneGraphEncodingNet\nets.py
Download dataset University-Release from University1652 (https://github.com/layumi/University1652-Baseline) to path ./data like this:
.\data
\University-Release
\test
\gallery_satellite
\query_drone
We provide a pre-trained model in ./weights
The R@5 recall results will be saved to file ./recall_results.csv