Lecture 04 - Interest Point Detection

[@hallpaz]

Thank you! You are helping education worldwide! - by a Brazilian student

[@afterbunny257]

The first time I tried to understand harris corner, it cost me a few days and still out of blue. Now it is very comprehensive, thank you sir!

[@helloansuman]

He is the best... Exactly the whole series. Please upload more such research oriented concepts. Very good for researchers starting their research. Thank you.

[@farihazulfiqar]

Hi Mr Ansuman, may I know what is your research area?

[@arminhejazian5306]

Thank you, this video helped me understand the math behind harris.

[@squirrelbrains2197]

Thank you for posting these videos. I was trying to learn the contents from slides that just contain mathematical formulas without explanations. I was very lost. (a German computer science student)

[@vijaynidumolu1941]

Thank you Dr Mubarak Shah, for sharing videos in youtube

[@kaustavisi]

For the first time in my life feeling grateful to a Pakistani-- A poor Indian student who doesn't have much exposure.

[@manub.n6223]

+Minus Infinity Same here man, this guy is a very good teacher.

[@umertariq4370]

whatever you do , whoever you become (as a teacher) a shade of being a Pakistani teacher always go with you. not denying that the overall video is awesome but jis tarah kuch baton ko slides dkh k pela hai

[@pervaiz.khan1857]

As a part of my Ph.D. course, I studied one subject here from a Professor of a German university. He also does the same !!

[@msrizal2159]

Thank you Dr Mubarak Shah, Allahu yubarik fik.

[@zahraaalmhana9157]

Can you get the source of this information. Please Help

[@tommyfan6911]

Your lecture saved my ass from my mid term exam.

[@pedropeter_]

That is one of a GOOD lecture.

[@shivangitomar5557]

Amazing!! Thank you!

[@rahimpanahi2302]

Thanks for this valuable collection of Machine vision lectures ......it helped me alot

[@zahraaalmhana9157]

Can you get the source of this information. Please Help

[@احمدالمليجي-ه8ت]

thanks professors , but why you used taylor series

[@akshayrao1484]

what is the dimension of the M matrix ?

[@inzar86]

if there is a block diagram of interest point detection process?

[@soumiasoumia3921]

thanks, how did you calculate the derivative of I(x+y,u+v) please 23:44

[@rohetoric]

Taylor series I(x+u,y+v)=I(x,y)+differentiation of inside wrt x u*I(x+u-x)/dx+differentiation of inside wrt to y ie v*I(y+v-v)/dy

[@zahraaalmhana9157]

@@rohetoric Can you get the source of this information. Please Help

[@rohetoric]

@@zahraaalmhana9157 Its Taylor series.. You can google it and check.. Its a common formula..

[@miptPatriot]

Hello. Thank you for lectures.
There is an error in description. Presenation link points to lecture 3. The right link is /courses/CAP5415/Fall2012/Lecture-4-Harris.pdf

[@rabiafyz9952]

Please i am very much interested to know about threshold to detect corners. How can we set threshold value to detect corners?

[@AbdullahAkay]

very good lectures

[@icelake526]

what a good video! Thanks a lot!

[@sahilmakandar773]

Where I find SURF algorithm mathamatical implementation

[@zahraaalmhana9157]

Can you get the source of this information. Please Help

[@ahmadslo7238]

thanks very much; it is so helpful

[@samyu4473]

In the last slice, it says "convolve these three images with a large Gaussian (window)", what do you mean by large Gaussian window, is it in terms of window size or sigma of gaussian, also, what is considered large in practice? Thank you

[@samyu4473]

never mind, since large sigma indicates large window size, i.e. should approach to 0 at the edge. The Szeliski book says it works best when sigma=1 for the derivative gaussian and sigma=2 for integration gaussian (gaussian used for window function w(x,y))

[@manjulagururajh]

Hello Sir
Thank you for the useful information about interest point detection

[@zahraaalmhana9157]

Can you get the source of this information. Please Help

[@malharjajoo7393]

How is 27:43 equation of an ellipse ?!

[@Globalvibes]

20:42 what if there comes noise. then the selected point as point of interest will not be visible. and we have discussed that it should be noise robust.

[@soumiasoumia3921]

Waseem Ullah did you know how he calculated the derivative of I(x+y,u+v) please 23:44

[@Globalvibes]

yes. that was easy. I(x,y) was canceled with -I(x,y). and wrote the remaining term

[@soumiasoumia3921]

Actually, I'm talking about the shifted intensity term how did he get its derivation using Taylor's formula I want to know even it is easy for you !!

[@php5engineer]

+soumia soumia, Wikipedia article on Corner Detection has much more elaborate description

[@waniubaid7718]

How to get his slides plx help

[@paladin1410]

Can anyone explain why we can use Gaussian window there? I think that a uniform window is enough because we just use it as a kernel to look for a corner reason in our image. In my opinion, the gaussian window is some kind of weighted average one which may affect on our algorithm.

[@aanchalharit]

thanks for uploading.

[@malharjajoo7393]

I think explaining the formula of the Corner Response would have been helpful. Many times just assuming formulas does not help convince the viewer.

[@aakashnandrajog7035]

What is K & L in correlation?

[@anguramesh1683]

thank you

[@leetcode_hardcore]

What's the full link??

[@abd_alkader]

hello can any one help me about >>> why ( if the result of correlation is high so there are similarity ) i think it is possible to be high even if there is no similarity

[@trsteng1088]

Thank you so much~

[@rabiafyz9952]

R > 0, does not bring only real corners. it also include fake corners. so how can it say R > 0 for corners?

[@leetcode_hardcore]

Thx ;)

[@biangbiangtv3167]

顾得

[@biangbiangtv3167]

三克油