Compile OK, result not ....well

[Xinghy98]

Hi， Daniel
I am interested in your multiH project very much, I have read your related paper many times. When i tried to compile your project, i can not get the same result picture as your paper, that is to say, whether the SIFT keypoint or mean-sift, i think maybe somewhere i do wrong.i hope you can instruct me how to realize the good results of pictures like where your paper shows

[Xinghy98]

@danini
Hi Daniel, I run the multi-H project with the pictures you show, but I can not get the same result picture as your paper show. I find the extracted SIFT keypoints are not the ketpoints as you have extracted, my results show almost corner points, however, your result in paper seems even in the plane. I want to know which means you use to extract keypoints or cluster planes? Thanks for read my question.

[danini]

Hey! It is a quite old code, give me a few days to see what happens!

[danini]

To me, it seems to be working well. I was trying to run the executable. Have you tried compiling yourself or you used the executable?
Regarding features:

• In the paper, the reported images/results are on the points provided by the dataset with additional affine frames detected as I remember.
• In the code, ASIFT is run to make it applicable to all image pairs not just the ones in the datasets.
  You can of course any time replace ASIFT by any affine-covariant detector (Hessian-Affine, MSER, etc.). But SIFT won't work since it does not provide affine features.

[Xinghy98]

@danini Hi Daniel, Thanks for your response! It is my honour to receive your response! I write this letter as soon as I see your message. I have compiled the project myself and run it in Linux, but I have not used the executable you provided because of the environment. The input picture is choosed from the dataset you provided. ASIFT has computed the Affine matrix between pictures. I can get the same picture as the one in the folder named result. I wonder how to get the same pictures as what your paper showed. The config is kept the same as your code show(the parameter named "method" in the function named process is "SIFT". whether the results in your paper use the "SIFT" config?). I put my running result in the follow. The Affine frame means what? I don't understand well, I know that you use HAF to solve Homographies matrix for each point . Thanks for read my question! I am very happy to wait for your response! I hope you can make more and more great progress in plane extracting and have happy mood everyday! Thanks for your response again! I admire for your great achievements!

[Xinghy98]

@danini

[six-finger]

Hi @Xinghy98, May I ask you a favor?
Could you please share me the method of modifying and compiling the project on Linux, as you said you've run it successfully?
I would greatly appreciate it if you could offer some assistance.