search

ubc-vision / image-matching-benchmark

Public release of the Image Matching Benchmark: https://image-matching-challenge.github.io
https://image-matching-challenge.github.io
Apache License 2.0
525 stars 93 forks source link

how to evaluate on my own dataset? #48

Open biggiantpigeon opened 1 year ago

biggiantpigeon commented 1 year ago

I would like to evaluate various features on my own dataset, I read the parts of the theory of generating scene data, but I cannot find the script to do this, and I don't know how those .h5 files are made. Any clue of how to do this? Thanks.

ducha-aiki commented 1 year ago

Originally we didn't intend to run the scripts with other datasets. However, you can check this instructions: https://github.com/ducha-aiki/creating-data-for-imc

If you don't use reality capture, skip that part