AI is just around the corner.

Started 4 months ago | Discussions thread
fferreres Senior Member • Posts: 2,724
Re: Which computation do you want?

Wayne Larmon wrote:

stevo23 wrote:

Steve Bingham wrote:

The future is going to see much more computational and AI logic in our cameras. In fact, I think Fuji, Nikon, Canon, et al are working on this right now. Where are the camera companies? Heck, smartphones seem way ahead - possible because of a MUCH, MUCH bigger market. Stay tuned. The camera market will either change - or die. the camera you own today will not resemble the camera of ten years from now. Ten years! Probably a lot sooner.

It's always helpful to talk about what it is you want in computational logic.

  1. Intelligent, robust frame stacking that works at hand held shutter speeds, with some subject or camera motion. Google Pixels do this. Not sure about other smartphones. Pretty sure no existing ILC or P&S does.
  2. Fake shallow DOF as good as most flagship smartphones can do. (Where the depth map can be repainted in PPing.) Yeah, my bokeh standards are low.
  3. Near perfect auto-metering so there is (almost) no need whatsoever for a histogram.
  4. Near perfect auto tone mapping so there is no need to shoot raw (most of the time.) (Again, no existing ILC or P&S can do this.)

My Pixel 2 has 1, 3, and 4. I don't think the Pixel 2 is best-of-breed with shallow DOF emulation. For me, 3 and 4 are "100% of the time" with my Pixel 2. "Extremely rarely" with any of my Canon cameras so I shoot 100% raw and plan on spending minutes per image duplicating what my Pixel 2 does automatically with SOOC JPEGs.

Also, in journalism and sports, I suspect much of it may not be allowed.

Agree. The above don't work at high FPS rates. But if implemented on an APC-S or FF sensor the worst case is the AI isn't used at all so the cameras work as well as they do now. Nothing is lost.

The Pixel AI is very robust. It degrades gracefully when fast FPS are used. It (almost) never causes image problems; you just get noisier images when the AI can't keep up.


Good! It’s like instant food. A good instant meal Can be goood. Even gourmet. It’s impressive how they can make tiny roost sensors do so much by exploiting computer vision and other algorithms.

What everyone needs to accept is that most camera makers retreat to niches, while camera makers expand based on only what most wants. Want the best legendary bokeh you can have? Like a $1000 tuna dish, you may want to look at Leica 70mm f1.4 made 60+ years ago. What bokeh that will delight and satisfy 99.32% of camera users? Use the phone Portait mode.

The more camera makers offer MORE than what this group needs, the more they waste their money. The more that camera makers focus on niche, the smaller they become and the faster they are outinnovated. They can only survive so long their niche is so small as to be uninteresting to others.

Do I believe a phone competes with my lenses? No, because I have needs/wants outside what the camera makers now cover. Because a small sensor while exploited to the max, still isn’t my nirvana.

However, the point that a phones does a huge lot better in terms of AUTO mode and how it extracts unbelievable amount of quality out of those little sensors, is unmatched for the general case. Even video has tremendous performance.

Phone cameras are like a killer swarm of mosquitoes, dressed like Cinderella, taste like strawberries and you can’t chose to have or not have them: they can’t be excluded from any modern top line phone anyway.

Post (hide subjects) Posted by
(unknown member)
(unknown member)
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow