Just me having a bit of weekend fun...in a positive way.
I'm so excited about the prospects of a resurgence in ML development, fingers crossed.
I see AI based ideas cropping up all over the place, eg this latest one from Leica https://youtu.be/i2juNxPsHe8
You see these things unfolding in other cameras, and your mind opens regarding what ML could do in a modern Cam, eg an R series, with Lua scripting.
Either I'm missing the point or I am missing the point.
Why is this something to write home about?
You still have perspective correction to take place and your image will be cropped accordingly and content will get lost in post. In worst case your composition (the idea behind the photo) will get destroyed. Nothing new so far.
Garry was just spreading some optimism :)
@names_are_hard
Thank you ;) :)
It's becoming impossible to post on this forum because people only look literally at things.
Oh well, such is life when you have an international forum :) ;)
I like this place and different viewpoints and personalities. That is, reactions and questions should be encouraged. Keep an open mind etc. A lot of personality doesn´t mean it´s always has to be personal ;).
Besides, I think we know each other quite well by now 8).
A 11min video to explain a feature that could have been explained in 30 seconds...
But yeah, I get your point garry. I just don't see how ML can get better, it's kinda hardware-limited by now.
While we are at it, I've been testing frame interpolation using RIFE (https://github.com/hzwer/arXiv2020-RIFE) and I immediately thought about how this could benefit low-framerate/high-resolution MLVs. For example, recording at 12fps 4k and post-processing using RIFE to get 24fps.
Another idea: people are using canon cameras to build datasets to train machine learning. Magic Lantern could help to annotate this dataset, adding metadata that could be manually selected on-the-fly. For example: you could input 10 keywords. After each photograph is taken, ML asks what keyword to pick and automatically writes to the metadata that can then be parsed on the computer.
I have also been testing with video super-resolution, but there's not much future for amateurs because networks like EDVR/TecoGAN requires too much VRAM. Only ESRGAN is possible at the moment, but it has no temporal consistency so it's not really usable for video (although the results on photographs are amazing). There's also denoising, where DPIR works very well.
So, indeed, machine learning software can really help with the hardware limitations that (old) canon cameras have.
QuoteI just don't see how ML can get better, it's kinda hardware-limited by now.
We all look at things through different eyes ;D
The video was not the point of my post ;)
I see untapped potential, ie building on the shoulders of giants.
Putting ML into new/modern Canon cameras: obviously.
Enhancing Lua scripting to give users what they individually want/need.
Bottom line: in the camera world, IMHO, ML is unique, and the star feature is scripting :)
I hope I don't interrupt you. I just wanted to mention that the Blackmagic design idea also was a career changer. It isn't only the potential. Philip Bloom said, in his earlier YouTube videos. "I am sure Magic Lantern engineers are working on it this minute". Maybe all that work and success wasn't payed fairly. In the future of course it would be possible using Magic Lantern on the R5/R6. Having a proper strategy would maybe open the doors for other projects and it wouldn't seem like some stupid nerds wasting their time. The future can't be Rocket Lab (source: pcwelt.de) using a 5d3 for pictures of the earth by the way without Magic Lantern.
In a science fiction world, magic lantern users would have easy access to crazy external hardware,
like this sd card with a "virtual gimbal" intergrated:
https://petapixel.com/2016/08/11/sd-card-built-gyro-sensor-stabilize-shots/
https://github.com/yossato/virtualGimbal
Maybe, we could even attach a smartphone over a camera and use its gyro sensor to stabilize footage