Are there any plans to add sift and surf ??

[juliaimages]

Are there any plans to add sift and surf ??

[tejus-gupta]

SIFT and SURF are patented [1], [2] and hence can't be used freely (for commercial purposes; research is fine) without permission. I am not sure whether they can be included here. What do you think, @timholy?

FREAK and BRISK are excellent alternatives to SIFT/SURF and are available in ImageFeatures.jl.

[timholy]

They can't be in this package because it has an MIT license, and that isn't going to change. However, one could start a SIFTFeatures.jl or PatentedImageFeatures.jl 😉 package with a different license. Contributions welcome!

[juliohm]

So the algorithm was patented? Interesting I didn't know that. I will avoid SIFT in the future.

…

On Fri, Dec 14, 2018, 07:42 Tim Holy ***@***.*** wrote: They can't be in this package because it has an MIT license, and that isn't going to change. However, one could start a SIFTFeatures.jl or PatentedImageFeatures.jl 😉 package with a different license. Contributions — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#59 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADMLbT4Z40f_T52SPGJLn0pFzfAk3QbWks5u43KNgaJpZM4ZRzj-> .

[zsz00]

ok, I See.

[timholy]

Should we add this information to the docs or the README?

[zsz00]

I think there should be more information in the README .

[BASS10]

So as far as I can tell, the SIFT patent has now expired (in the time since this thread was last active). Is there anything else standing in the way, or could SIFT now be added to ImageFeatures? I would be interested in helping in any way I can. I'm fairly new to Julia though so I can't vouch for how much use I would actually be, but the offer is there.

[timholy]

Thanks for the offer! Can you supply a reference for the expiration of the patent?

Does SIFT offer advantages over the feature-detectors already here? (I'm aware that it's widely used.)

[BASS10]

Thanks for replying so quickly. The patent reference above from @tejus-gupta now shows the status as expired.
https://patents.google.com/patent/US6711293

I'm interested in SIFT for photogrammetry, where it seems to be one of the standard choices.

Having had a brief play with the lighthouse examples I've found the existing feature detectors don't seem very robust for scale transformations. Perhaps adjusting the parameters would help, I haven't tried much of that yet. I'm hopeful that SIFT might help.

Since the patent has expired, it would be nice to have an implementation in Julia and this seems like a good place for it to live.

I will try to get my head around the existing codebase in the meantime; but no promises! Any direction would be welcome.

[kmsquire]

At the time SIFT came out, it was groundbreaking, and it was better than most descriptors that came after it for quite a while, extensions to SIFT itself notwithstanding.

I didn’t follow it closely after a while, and I haven’t kept track of what’s in this package, but my understanding is that other descriptors eventually caught up, although I haven’t looked for/at any recent comparisons.

I think it would be nice to have SIFT here. It would be nice to include in comparisons, and there were some nice extensions that (I think) were not patented, and that could also be included.

That said, I don’t personally have time to contribute any of these, beyond giving references (so thank you to @BASS10 for volunteering!).

[orkolorko]

@BASS10 I'm interested in helping

[jenkspt]

Here's a pretty straightforward implementation of SIFT in pytorch.
https://kornia.readthedocs.io/en/latest/_modules/kornia/feature/siftdesc.html#SIFTDescriptor