image-matching-toolbox robotcar-season eval run very slowly

Hi, thanks for your excellent work.

I finished robotcar dataset preparation and can successfully run eval program. However, the process takes too much time (around 2000+ hours estimation for matching 4.7M+ data). Is it normal? Or any mistakes maybe happen?

Also, does skip-match work? And can I skip that? Could you help me and explain?

Thanks for your time!

Jan 14 '23 09:01 XJTU-Haolin

Hi! I'm facing the similar problem, have you solved it?

Feb 18 '23 11:02 autostudiolea

Hi @XJTU-Haolin ,

I think the easiest way to check this is by estimating how long your target matching method takes to match a single pair. And then you can compute how long it requires to eval 4.7M+ data. To quickly evaluate the runtime of your matcher within your environment, you can use the example pairs or hpatch sequences. And make sure you are using GPU when you run the matcher. It should be using a GPU by default as long as it is visible in your environment.

Feb 27 '23 13:02 GrumpyZhou

Hi @XJTU-Haolin ,

I think the easiest way to check this is by estimating how long your target matching method takes to match a single pair. And then you can compute how long it requires to eval 4.7M+ data. To quickly evaluate the runtime of your matcher within your environment, you can use the example pairs or hpatch sequences. And make sure you are using GPU when you run the matcher. It should be using a GPU by default as long as it is visible in your environment.

Thanks for your excellent repo and nice response. I will find time to check again.

Mar 28 '24 12:03 XJTU-Haolin