It's pretty fitting that a lecture about compressive sensing is uploaded in 240p. I assume reconstructing the video quality is "left as an exercise to the viewer"
I bet even in the year 2100, somebody will watch this video and think "Isn't it just beautiful". One of those talks that leaves you enlightened at the end.
The camera focus should be on the slides rather than on the speaker. It is bit annoying at times, when useful information is shown on slides and camera is focused on the presenter.
fantastic talk. Unfortunately i did not understand how to reconstruct. Since the recovery of the sparse signal is sparse in a unknown basis... so how do i know which basis to use so i can reconstruct?
I think they addressed this in discussions. So you need to decide on an appropriate basis based on domain knowledge, for eg. some images would be best recovered using a wavelet basis. In the future, if someone comes up with a better basis, you can use that to recover better quality images. This was one of the questions asked.
Have a look at databookuw.com/ There is an entire chapter and Jupyter Notebooks on compressed sensing. Also, the notebooks are sometimes extremely slow because they are using scipy.optimize. As soon as you get frustrated with waiting, have a look at cvxpy for a fast convex optimizer lib.
It's pretty fitting that a lecture about compressive sensing is uploaded in 240p. I assume reconstructing the video quality is "left as an exercise to the viewer"
I bet even in the year 2100, somebody will watch this video and think "Isn't it just beautiful". One of those talks that leaves you enlightened at the end.
The camera focus should be on the slides rather than on the speaker. It is bit annoying at times, when useful information is shown on slides and camera is focused on the presenter.
fantastic talk. Unfortunately i did not understand how to reconstruct. Since the recovery of the sparse signal is sparse in a unknown basis... so how do i know which basis to use so i can reconstruct?
I think they addressed this in discussions. So you need to decide on an appropriate basis based on domain knowledge, for eg. some images would be best recovered using a wavelet basis. In the future, if someone comes up with a better basis, you can use that to recover better quality images. This was one of the questions asked.
en.wikipedia.org/wiki/Sparse_dictionary_learning
The basis is always known, you get to choose it, that's the point.
do you have any document for this talk?
Hey, this textbook aligns somewhat well with what he is talking about www.ecs.umass.edu/~mduarte/images/IntroCS.pdf
Thank you =)
I would like to program it in python. Can anyone suggest on this?
Have a look at databookuw.com/ There is an entire chapter and Jupyter Notebooks on compressed sensing.
Also, the notebooks are sometimes extremely slow because they are using scipy.optimize. As soon as you get frustrated with waiting, have a look at cvxpy for a fast convex optimizer lib.
@@nils3030 Great, Thank you!!!😊
@@nils3030 I need the python code, is it available?
Really cool
Is everybody with me?