seam_carver

This might be relevant. I took fullscreen screenshots. The first one was 2560x1440. The second one was 5120x2880.

I think it's specific to 1440p

This might be relevant. I took fullscreen screenshots. The first one was 2560x1440. The second one was 5120x2880.

On a related topic, people seem to not like me using Medium. What platforms do you recommend?

My specific issue was it was still hot and noisy even when the built in display was closed. I'm sure other people have 1440p monitors and would love this partial solution.

Sorry, I just defaulted to using Medium, didn't realize that people would have trouble reading it! You can still open it using incognito mode to read it. Or use the non paywall link https://axu2.medium.com/solved-by-1440p-hidpi-macbook-pro-16...

You can just open the link in incognito mode! Also, sorry, I guess I should have posted the friend link instead that doesn't ask to subscribe, but I can't change it after posting! But here's the non paywall link: https://axu2.medium.com/solved-by-1440p-hidpi-macbook-pro-16...

I just quoted the 200 page MacRumors article with the same clickbaity title, since I though people might recognize it.

Just open the link in incognito or something. Or click non paywall link https://axu2.medium.com/solved-by-1440p-hidpi-macbook-pro-16...

Or they are just doing this intentionally to make M2 look good 0.0

If you can't open the link, use incognito mode! Or click the non paywall link: https://axu2.medium.com/solved-by-1440p-hidpi-macbook-pro-16...

This issue has been plaguing me for ages and I only just now figured out the solution!

Also I should just write this in Cython since I do a lot of unnecessary stuff when using numpy.

After talking with some friends, basically I could get this to be a lot faster if I just wrote it in Cython. (the skimage.transform.seam_carve function is written in Cython). Also, I realize now that the double for loop in Cython is actually fast, so that means that the skimage implementation of dynamic programming is optimal. But I can still change it to accept the accumulated energy matrix!

That's a lot to take in; thanks for all the links!

That looks pretty cool! Thanks for the link! I'll look more into it soon.

My goal right now is to get as close as I can to Photoshop speed; I'm fairly certain that the source code for Photoshop is not available online...

I actually realized I was doing a lot of unnecessary computation in my dynamic programming step. I removed it and from 14 s it's now 8 s.

It's 14 seconds to remove 200 seams on the 512-by-342 bench image, but I note potential optimizations in the notebook.

Right now, the seam_carve function in scikit-image accepts an energy matrix, which it then uses to generate the accumulated energy matrix. Forward energy calculates the accumulated energy matrix directly, but I calculate the energy matrix backward from it to be able to use scikit's seam_carve function.

If I write my own seam removal function or change scikit's, I think I could be able to make the speed comparable to backwards energy.

Furthermore, the accumulated energy matrix calculation step in scikit uses a double for loop, which might be able to replaced with some np stuff for a speedup for both forward and backwards energy.

I should probably raise this as an issue on GitHub... thanks!

By GAN, do you mean "Generative adversarial network?" I don't have much ML experience so I don't really understand what you're saying :P

Do you know a good place to start? I heard Andrew Ng Coursera course is good.

I don't know most of the terms you just said... I guess I should try learning some ML? I heard the Andrew Ng Coursera course is a good place to start.

That assignment only covers the original seam carving algorithm; this post also covers the improved seam carving algorithm!

I'm unfamiliar with CNN's, but you can draw masks manually to protect/remove parts of an image, or use a face recognition thing, or a motion detection thing, etc.

Don't think most people know about the improved seam carving algorithm (since only the original is taught in most CS curriculums) and thought you might think this is cool!