trisa's Introduction

TriSA: A Three Stage Approach for UAV-Satellite Cross-View Geo-Localization Based on Self-Supervised Feature Enhancement

This repository is the code for our paper TriSSA: A Three Stage Strong Approach for UAV-Satellite Cross-View Geo-Localization Based on Self-Supervised Feature Enhancement, Thank you for your kindly attention.

requirement

Download the University-1652 dataset
Prepare Data Folder

├── University-1652/
│   ├── readme.txt
│   ├── train/
│       ├── drone/                   /* drone-view training images 
│           ├── 0001
|           ├── 0002
|           ...
│       ├── street/                  /* street-view training images 
│       ├── satellite/               /* satellite-view training images       
│       ├── google/                  /* noisy street-view training images (collected from Google Image)
│   ├── test/
│       ├── query_drone/  
│       ├── gallery_drone/  
│       ├── query_street/  
│       ├── gallery_street/ 
│       ├── query_satellite/  
│       ├── gallery_satellite/ 
│       ├── 4K_drone/

Evaluation and Get the Results in Our Paper

You can download the trained embedding files (.mat)from the following link and put them in the "evaluaiton/weights/" folder

Download the trained files

Google Driver

You can download the trained embedding files (.mat) from the following link and put them in the "evaluaiton/weights/" folder

We prepared the following .py files for easy evaluation:

├── evaluation/
│   ├── evaluate_cpu_no_rerank.py  /* test with cpu and no rerank
│   ├── evaluate_cpu_rerank.py     /* test with cpu and rerank
│   ├── evaluate_gpu_rerank.py     /* test with gpu and rerank

If you are using the gpu-based re-ranking, make sure to compile the file by:

cd evaluation/
sh make.sh

Train and Test

We provide scripts to complete TriSSA training and testing

Change the data_dir and test_dir paths and then run:

python train.py --gpu_ids 0 --name traied_model_name --train_all --batchsize 32  --data_dir your_data_path

python test.py --gpu_ids 0 --name traied_model_name --test_dir your_data_path  --batchsize 32 --which_epoch 120

Thanks

Zhedong Zheng, University-1652: A Multi-view Multi-source Benchmark for Drone-based Geo-localization
Xuanmeng Zhang, Understanding Image Retrieval Re-Ranking: A Graph Neural Network Perspective

trisa's People

Contributors

Watchers

trisa's Issues

resnet50_ibn_a.pth.tar

Thank you very much for your excellent work, which has inspired my research.I will thankful if you provide the resnet50_ibn_a.pth.tar

Train. py and mask. sh not found

Thank you very much for your excellent work, which has inspired my research. But the train. py and mask. sh mentioned in your readme were not found in the project, and I am very much looking forward to your provision

Recommend Projects