kregmi / cross-view-image-matching Goto Github PK

Hello, my name is Zhao Hu. Recently, I am researching the direction of cross-view geolocation. I read your paper and think it is very helpful to me. I would like to use the CVACT dataset. Could you give me a link, thank you very much

Thanks for sharing the OP dataset.

When I was using the dataset, I found that each line under the splits/train_split.txt file points to 4 files,
for example, 28.5366_-81.3741_nadir.png,28.53663_-81.37406.png,28.5366_-81.3741.png,28.53663_-81.37406.png.
Whereas each line in splits/test_split.txt points to only two files per line,
For example, 40.44_-80.0005_nadir.png,40.44001_-80.00056.png.

How are the four files in each line of train_split.txt trained? It seems that the third file in each line of train_split.txt does not exist, while the fourth file has the same file name as the second file.

Can you share the segmentation maps of the UCF-OP datasets? Thanks a lot.