Talk:Speeded up robust features

OpenCV[edit]

If OpenCV does indeed include a SURF implementation, the link to it should be made more clear. The existing link just links to the main OpenCV sourceforge page. The OpenCV wiki at http://opencv.willowgarage.com/wiki/SummerOfCode indicates that SURF support was planned, but I'm not positive it made it in Electron100 (talk) 21:43, 4 September 2009 (UTC)[reply]

---→ I updated the OpenCV link to be directly to the online documentation of the SURF function: http://opencv.willowgarage.com/documentation/cpp/feature_detection.html#surf User:Shervinemami 222.127.196.140 (talk) 06:08, 19 October 2010 (UTC)[reply]

Speed[edit]

This article is brilliant in a way that's rather spectactular, claiming both:

"The standard version of SURF is several times faster than SIFT"

and

"SURF has been shown to have the same performance as SIFT." (with citation)

Now which is it? Gijs Kruitbosch (talk) 13:29, 16 January 2009 (UTC)[reply]

"Performance" in this context means how well it performs, i.e. finds matches between points. --Petter (talk) 20:31, 22 January 2009 (UTC)[reply]

By this definition of "performance", the performance of SURF is far from better than the one of SIFT (See the referenced comparison). According to the referenced article [1] , SURF produces more correct matches per second than SIFT, while SIFT produces a higher number of correct matches per image pair than SURF. I am therefore removing the inaccurate sentence SURF has been shown to have a superior performance compared to SIFT. Sbstn (talk) —Preceding undated comment added 12:38, 15 April 2009 (UTC).[reply]

Eternal links[edit]

Kind of curious why the VisionLab link was removed from the implementations. I thought this section was intended for links to SURF implementations and VisionLab implements SURF. —Preceding unsigned comment added by Bmitov (talk • contribs) 00:08, 2 January 2010 (UTC)[reply]

those were filthy competitors :) 95.133.221.200 (talk) —Preceding undated comment added 00:09, 20 February 2010 (UTC).[reply]

Cleaned up external links section and formatted it into a table. It should contain all the original information but be much easier to read. If anyone has some time please fill in the blank parts because I was not sure what should go there.

Added a section on performance studies, which was needed considering how many options there are. One of the links goes a study that I did. If anyone has an issue with this self referral please comment here. --Pabeles (talk) 16:06, 15 November 2011 (UTC)[reply]

To address the overuse of external links tag, I removed the list of implementations per guidelines on external links and what Wikipedia is not. I also removed the section on performance studies per the conflict of interest guideline. - MrOllie (talk) 15:43, 29 November 2012 (UTC)[reply]

notation[edit]

This expression appears in the article:

\sigma _{\text{approx}}={\text{current filter size}}\times \left({\frac {\text{base filter scale}}{\text{base filter size}}}\right)

Aside from neglect of proper indentation per WP:MOSMATH, it said this:

\sigma _{\text{approx}}={\text{current filter size}}*\left({\frac {\text{base filter scale}}{\text{base filter size}}}\right)

(except that in addition, every initial letter was capitalized). That is vulgar. Using an asterisk for this purpose is a workaround for occasions when one is restricted to the characters on a keyboard. One can write

5\times 3{\text{ or }}5\cdot 3.

Michael Hardy (talk) 18:39, 31 July 2017 (UTC)[reply]