search

gilbaz / LORAX

CVPR Paper - "3D Point Cloud Registration for Localization using a Deep Neural Network Auto-Encoder" - Partial Implementation Code
91 stars 29 forks source link

Code for SAF and coarse matching #1

Open brejchajan opened 7 years ago

brejchajan commented 7 years ago

Hello, this is more a question than a real issue. First of all, thank you very much for a great paper. I would be interested in testing your algorithm, however, as I can see only the part for creating RSCS is available. Do you consider releasing the rest of your code and models for coarse matching in the near future? Thank you very much!

gilbaz commented 7 years ago

Hi,

Thank you very much.

Yes, so right now only the RSCS is available. The algorithm was mostly written in Matlab, so I am trying to translate it to python for a fast and more efficient version (that will be posted to GitHub). I can't release the Matlab version, not high enough quality for public open source use)

The SAF is pretty trivial with TensorFlow (exactly how I described in the paper)

The matching has a bunch of little coding nuances but algorithmically it is also exactly like the paper.

Sorry, I really wanted to finish it before CVPR but I just am super busy. I will try to do it in the next month - two months. In any case if you have questions or suggestions feel free to ask.

Thanks, Gil

On 18 Aug 2017, at 23:34, brejchajan notifications@github.com wrote:

Hello, this is more a question than a real issue. First of all, thank you very much for a great paper. I would be interested in testing your algorithm, however, as I can see only the part for creating RSCS is available. Do you consider releasing the rest of your code and models for coarse matching in the near future? Thank you very much!

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

tanajikamble13 commented 5 years ago

@gilbaz Can you release the rest of code. I am trying to thoroughly understand your paper. If you release rest of code then it will help me a lot. Thanks, Tanaji

gilbaz commented 5 years ago

Hi,

Sorry, currently I'm not able to get enough time to rewrite the original code. In general I highly recommend attacking this problem with newer methods that may be more robust that the very basic network that I used. Try and define your super-points in a way that will provide you with maximum geometric information and then this problem is reduces to a combinatorial RANSAC based problem with well known solutions.

Best, Gil

On Wed, Nov 14, 2018 at 3:50 PM tanajikamble13 notifications@github.com wrote:

@gilbaz https://github.com/gilbaz Can you release the rest of code. I am trying to thoroughly understand your paper. If you release rest of code then it will help me a lot. Thanks, Tanaji

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/gilbaz/LORAX/issues/1#issuecomment-438667986, or mute the thread https://github.com/notifications/unsubscribe-auth/APupAFzYZw5RJmGH2NF-kLE1HCzet51iks5uvB-UgaJpZM4O8A9D .

tanajikamble13 commented 5 years ago

Thanks for immediate reply. I will follow your recommendations.

siliangwu commented 4 years ago

@gilbaz Hello, Can you release the rest of code now? Some of my master's thesis experiments need to borrow your method. Thanks, Wu