Photogrammetry II - 10 - SIFT Features and RANSAC (2015/16)

[@chotnt]

The approach to introduce the idea is awesome!!

[@sarazhalehpour1283]

The best lecture I have seen for SIFT features so far.

[@CyrillStachniss]

Thanks!

[@shiqiai2881]

exactly. the best~

[@mohammadaminmousavi5011]

The BEST and the most helpful lecture for SIFT and RANSAC that i ever seen. Thank you Prof.Stachniss

[@W00PIE]

Being a 40+ dev who's never seen a university, I started with CV two days ago because of an urgent and interesting project at work. Feature detection/matching seemed like the right thing to look for after some research. Now, after watching and digesting your talk, I feel extremely confident about how I'm going to tackle the problem. I'm building a non-interacting optical device monitoring system for about 3.5k rail infrastructure sites that is supposed to run on minimal ARM hardware using gocv. You really know how to get the key points across, thank you for sharing. I'll definitely take a look at your other videos. Cheers from Krefeld!

[@bobthemagicmoose]

Excellent lecture! I was familiar with some of the concepts, but this lecture really gave me a strong grasp of the terminology and how these concepts interplay.

[@supundasanthakuruppu3496]

Thank you very much professor. It was very clear and I could understand the concepts clearly.

[@johnnysuzuki2908]

Thank you Cyrill! I benefit a lot from your clear explanation of SIFT and RANSAC.

[@alaamohammad2778]

you just simplified the paper of Distinctive image features from scale invariant keypoints, many thanks for you, you are amazing,

[@arunram6687]

This was good. Clear and the most intuitive explanation

[@zhaoxiao2002]

Motivation is explained very well. It is important for understanding. thanks.

[@stefano8936]

29:00 the two images are not at "slightly different point of views": it's actually the same. Easy to understand since all the keypoints are matching and all the lines are horizontal.

[@mohalemolefe]

Explanation here is top class! Thank you Cyrill👌🏽

[@TheGermanGuy91]

You really helped me writing my paper and understanding the concepts. Cheers mate.

[@Mlantow20]

19:16 thsmos, smos, smoos, EVEN MORE THSMOOST. Great lecture by the way !

[@AndreaCensi]

Nice lecture, Herr Doktor Professor! I watched this the night before the day I was supposed to give a lecture on the same topics :-)

[@CyrillStachniss]

Thanks Andrea, hope it helped. If you need the pptx slides, let me know. Cheers!

[@amnanajib8167]

For the sift, I was wondering if we take the original image and the one after smoothing and get different maxima, which points are then to consider as key points?

[@childhoodgames1712]

Thank you so much, ( But there are some equations are used at David Lowe 2004 paper describes the SIFT algorithm, could you talk about them to understand the application of those equations clearly? )

[@smazi00]

Nice lecture sir, can you suggest a way to apply SIFT to omnidirectional images?

[@fablungo]

Maybe I have misunderstood other lectures on it, but I think the way you describe how SIFT handles scale invariance (and even how you have annotated the diagram) at 19:50 implies it is the different size images that provide the scale invariance whereas my understanding is that it is the differing levels of smoothing that are finding gradients at different scales and that the different size images that are shown in the diagram is just an optimisation based on the fact if the maximum frequency is halved in each dimension (as a result of smoothing), then the image can be subsampled to half the size in each dimension without loss of information. Great lecture otherwise, though. It was very comprehensive. Thanks.

[@mauroboreggio835]

Hi fabrizio. Could you better explain the concept of scale invariance you are pointing to?

[@ivanperez7713]

Thanks, clear explanation of the RANSAC algorithm

[@elena_stamatelou]

Very helpful to understand practically the topic

[@Ub4ys]

thanks a lot for RANSAC explaination. it's really helpfull

[@aseelmsc2121]

Very helpful to understand RANSAC thanks a lot

[@TheTacticalDood]

Hi Cyrill, do you think it is still valuable to learn hand-engineered features such as SIFT in the era of CNNs and deep learning?

[@amnanajib8167]

How is e in real life given, I don't think so will sit there and count how many outliners are there?

[@manikabindal1392]

I needed to understand RANSAC...thanks a lot for the nice lecture..:)

[@dali7572]

Excellent explanation

[@1MacDuck1]

This was so awesome, Thanks!!!

[@yousefhajhamoud8854]

Thank you very much for this wonderful lecture.

[@iitansrocks5863]

sir plz help me using for matlab code according area

[@loneband302]

Until now, I still trying to find the value in keypoints using Opencv and python...hope anyone can help me.

[@polaw7204]

43:46 Ransac algorithm

[@vlogsofanundergrad2034]

thanks

[@henriqueramosricci1728]

Excelent class!

[@whasuklee]

Great lecture! Thank you so much again!

[@vitaliiwellplaied1366]

Very helpful lecture! Thanks a lot.

[@kutsalozkurt]

Awesome lecture, thank you so much

[@ianchik]

Great lecture. Helped me a lot. Thanks :)

[@roar363]

Nice lecture !

[@CyrillStachniss]

thank you

[@your_name96]

RANSAC from 43:00

[@OmarAbdelhamid--]

I don't usually comment on any stuff but WOW WOW WOW WOW

[@anushanramesh2344]

Thanks a lot sir!!! It was really helpful

[@asiamarri5939]

very nice lecture. May i request the slides?

[@CyrillStachniss]

Hi, You can download the full set of Powerpoint slides (with TeXPoint formulas) here: www.ipb.uni-bonn.de/html_pages_staff/CyrillStachniss/stachniss-photogrammetry-slides.zip Best, Cyrill

[@asiamarri5939]

Sorry for late reply. Yes I have downloaded. Thaink you so much for your prompt response.

[@jiongwang7645]

can I click more than 1 once the thumb up ? Great lecture !

[@iota1154]

why is there 4*4 histogram in 39:01....I just count 4.........

[@W00PIE]

There are four quadrants (as seen on the right side), but each quadrant is comprised of 4x4 fields (left image). So in the end 16 histogram fields are condensed into a single quadrant.

[@sumanthbalaji1768]

32:28

[@wajahatnawaz2145]

it is really helpful to understand the SIFT keypoint detection etc. I need some help can you share your email address.

[@baturgunl504]

legatus glaber