gmberton/VPR-datasets-downloader

Issue with Running format_tokyo247.py due to MAT File Structure

Closed this issue · 2 comments

Hello @ga1i13o @gmberton !

I recently downloaded the Tokyo 247 dataset and the corresponding MAT file using a link provided by the authors of NetVLAD for running format_tokyo247.py. Initially, the MAT file was named dbfnames.mat, so I renamed it to tokyo247.mat as expected by the script. However, upon executing format_tokyo247.py, I encountered an error.

Upon inspecting the contents of the MAT file I downloaded, I found it structured as follows:

['__header__']: bytes (76)
['__version__']: '1.0'
['__globals__']: list (0)
['dbfnames']: ndarray(object) (1, 75984)

Moreover, the internal structure seemed to be in the format:
['dbfnames'][0][0] = ['/work/exp/Shibuya_2014/Shibuya_mid/db_gsv/03814/381426.780_3946120.387/VIfhhHgWyU5FW26SNHrzGQ_012_000']
['dbfnames'][0][1] = ['
/work/exp/Shibuya_2014/Shibuya_mid/db_gsv/03814/381426.780_3946120.387/VIfhhHgWyU5FW26SNHrzGQ_012_030']

It appears that the structure of the MAT file I received differs significantly from what format_tokyo247.py expects.
Could you advise if the correct tokyo247.mat file for the VPR-datasets-downloader-main is available somewhere, or if there is a specific version of the MAT file that I should be using?

Hi, can you try with this tokyo247.mat file tokyo247.zip? You just need to extract the zip.

Thank you for your reply!
tokyo247.mat file successfully works!!