hpatches/hpatches-benchmark

HardNet result with matlab and python

aIR1000 opened this issue · 7 comments

Hello,

I have tested Hardnet's pretrained model of liberty with augmentation (HardNet+) with python implementation of HPatches. I was able to reproduce results of paper but when I tested with matlab implementation of HPatches to generate graphs I wasn't able to reproduce retrieval results (patch verification and matching results were correct) Can you please guide what I possibly be doing incorrect. Split is "full", I have checked. Does it has something to do with distractors?

Hardnetlibertywithaug (mAP with python implementation) = 69.66 (same as reported on paper)
Hardnetlibertywithaug (mAP with matlab implementation) = 60.64

Thanks in advance

Hello,

I have tested Hardnet's pretrained model of liberty with augmentation (HardNet+) with python implementation of HPatches. I was able to reproduce results of paper but when I tested with matlab implementation of HPatches to generate graphs I wasn't able to reproduce retrieval results (patch verification and matching results were correct) Can you please guide what I possibly be doing incorrect. Split is "full", I have checked. Does it has something to do with distractors?

Hardnetlibertywithaug (mAP with python implementation) = 69.66 (same as reported on paper)
Hardnetlibertywithaug (mAP with matlab implementation) = 60.64

Thanks in advance

I also encountered the same problem, can you tell me your problem solved?
Thank you every much!

+1, @ducha-aiki @lenck Could you please help us with the problem?

No matlab, no problem)

No matlab, no problem)

I want to use it to evaluate my descriptor it in a paper. Without using matlab, I don't know how to generate the figure.
Is there any other method?

Check the branch ench/matplotlib-fig.

We have python code there to plot the results.

However it isn't fully tested. If you find that it works fine, we can merge it into the master branch.

Check the branch ench/matplotlib-fig.

We have python code there to plot the results.

However it isn't fully tested. If you find that it works fine, we can merge it into the master branch.

Thanks for your help!
I generate figure correctly and I guess it would be better if you modify the readme.md. Since it does not mention we should modify the utils/config.py to evaluate other descriptors.

I'm glad it worked, since as I said it's not fully tested. Time is pretty limited now for everyone, if you want you can send a pull request and we will merge it.