I am looking for confirmation from Leia on the use of AI in LeiaCam photos.
Assume a photo is taken with LeiaCam on either a Lume Pad 1 or Lume Pad 2. The left and right images from these photo files can be extracted to make a SBS image. Is it correct to conclude that no AI was used in this resulting SBS image?
I don’t work for the company, but from what I understand it’s not just a standard 2 lens stereo image. It is doing some sort of processing to produce either a depth map or some sort of lightfield data. Whether that uses AI or not, I’m not 100% sure, but it does seem to be more complex than just 2 photos.
It depends what you mean by AI.
None of the pixels you see in each view are generated, and there isn’t novel view synthesis being done. So all the pixels are captured, or “real”.
However, we are running AI (more accurately, machine learning models) for the depth mapping, which helps us auto-focus the camera properly on the subject, and for convergence, which helps us adjust disparity at capture time of the 3D image.
Thank you, Nima. Here is some brief background on my request for confirmation.
The Photographic Society Of America (PSA) has the following sentence in the 2024 Exhibition Standards Conditions of Entry (Rules):
Images created in whole or in part by image creation software (frequently called ‘AI’ images) are not allowed.
Based on your response, I think that the SBS image would not violate this rule and could be submitted to a PSA sanctioned Exhibition.
Yes, definitely. The image is not “created” by AI, the camera sensors are put in focus by AI and the stereo image rectification (read: 3D cropping) is being done by AI.
I think a lot of the people making rules and trying to regulate AI don’t really understand it.
I don’t really know how an expert could define the difference in the auto-focus system on the FujiFilm FinePix W3 and the Lume Pad 2, other than one is software that runs on a CPU and the other is software that runs on a DSP. But the software we’re using COULD run on a CPU, just much slower. The end result is the same, which is that you focus on a subject in the image.
Also, every smart phone camera since maybe like 5 years ago (or more) does processing, to adjust the focus, optimize the resolution of cheaper sensors, color correction, etc. You can call this AI, but that is mostly marketing. You’d have to buy like a old school film camera if you really could not live with some digital processing.