[Winder]

I would really love to see some files from A9 and 5D Mark IV from you where the A9 files are much better. Again, I'm not a fan of DXO charts, but the dynamic range from 5D Mark IV is rated better than A9. In practice, there is no difference in recovering shadows. Yes, the A9 has a better af-c, but as all other cameras from all manufacturers will miss focus from time to time in restaurants with dim light, even with the famous eye af. Maybe with the upcoming update they will make A9 af perfect. And yes, A9 also has a little better high ISO than 5D Mark IV. But, if you resize the 5D Mark IV from 30mp to 24mp to match the A9 files, the noise will be similar.

Here is a print screen with 2 zoomed images taken with A9. It's my father in law in the images, dancind with his wife. He knew he is being photographed so he didn't moved too much. These are 2 images from a series of 3 consecutives images. If you look at both images, you will see that on the left image the focus is on the nose, not on the eye like it is on the right. And I have quite a few similar images from that event where A9 missed focus. Yes, my 5D also misses focus from time to time, but it's far from being "average at best" to quote you.

I would say your experience with the A9 is definitely not normal.

To Sony From Canon: 8 Thoughts on 4+ Years of Shooting A7-Series Cameras - On Portraits

[Dan Rentea]

Winder, here is another example from where people can draw some conclusions. These 2 images are from a wedding were I was the godfather. The photographer hired to shot the wedding was using a Sony A9. He had 2 flashes on stands aimed at the white ceiling and one flash on his camera. I didn't had a flash with me because I was there to party, not to take pictures. But my wife wanted a image with her parents and I took a shot of them between 2 glases of wine. The official photographer was next to me and he also took a shot. I inserted in the print screen the exif for both images.

The official photographer image details are: Sony A9, f3.5, 1/200s, ISO 1000, 3 flashes (2 on stands and one on camera). Total cost of the gear (without flashes and stands): 6196$
Sony A9 body: 3998$ at B&H
Sony 24-70mm f2.8 GM: 2198$ at B&H

The details of my image are: 5D Mark IV, f2.8, 1/200s, ISO 2500, no flash. Total cost of the gear: 3648$
Canon 5D Mark IV: 3099$ at B&H
Canon 35mm f2 IS USM: 549$

Look at the details on my father in law wrinkles (without the help of the flash, as the photographer shooting with Sony A9 had). Do you think that if I was shooting at ISO 1000 instead of ISO 2500, with flashes, at f3.5 instead of f2.8 and with the 24-70mm f2.8L lens instead of 35mm f2, the Sony images would have been better? I really really doubt it.



But, people are influenced by Tony Northrup or DPreview reviews, they are also influenced by lab tests where someone shoot a paper with camera on tripod and with controled light and they actually think that in real life shooting conditions the differences in image quality between Pro full frame cameras are big enough to make a difference... The real differences are on:
- af (depending on what you are shooting)
- the fps and the buffer of the camera (depending on what you shoot)
- the system behind the camera (lenses and accesories)

And one major part (or influence) has the photographer behind the camera. Mirrorless will be the future, but for the moment the "average at best" DSLR cameras can challenge the high end mirrorless cameras anytime. The secret is to stop watching cancan reviews and go out and shoot and gain confidence in yourself.

[biz-engineer]

They're not accomplishing it in the way you think they are, not that simple.Google explains the magic behind the Pixel 3's Super Res ZoomGoogle AI Blog: Night Sight: Seeing in the Dark on Pixel Phones

I still do not see where they used Artificial Intelligence (A.I.) , except in the title of the article. They say that they trained the algorithm for the white balance, but then the algorithm is fixed, so by definition they aren't using A.I.

---------- Post added 16-02-19 at 18:45 ----------

"AI" is being co-opted by marketers as the buzzword of the month and being used in places where it really doesn't apply, at least not in the strictest sense. Computational photography is not AI.

Yes, you are right. Marketers now use AI everywhere even when they don't know what it is.

[surfar]

Do the current Zs have enough processing power in them?

Not sure?

However,the extra battery grip provides the power to give the Z series some incredible connectivity,the processor can handle that.

12 bit ProRes Raw exported via HDMI to an external recorder puts the Z6 into dedicated high end video bodies class @ the price(+ the accessories) of a consumer camera.Nikon dont have a line of video cams to protect like Canon.

[madbrain]

Aside from the under-whelming video specs, a few things about the EOS RP puzzle me:

What did you find underwhelming about it ? I don't see the bit rates.
It does 4K video, and more. The 30 minute limit is present though, which is a PITA.

One thing I found odd is that it only goes to 40,000 ISO in standard mode, 102,400 extended. While I would probably never use that high of an ISO, I have done great shot at 25,600 ISO on my K-1 II. I have had very poor noise results on my old Canon T3i .
The relatively low range for current camera generations tells me they might still have noise issues. Or maybe they are just being conservative.

Battery life: 250 shots
Connectivity: USB 2.0

250 shots is too low and a nuisance, plus USB 2 in this day and age is such a throw-back. Why would Canon bother to cut corners here?

250 shot is likely because of the screen having to be on all the time on mirrorless.

Agree the USB 2 is useless nowadays. You won't be using it to transfer RAW files, much less your 4K video files. They are just checking a box for their specs most likely. But they may as well remove it, if they are not going to upgrade it to 3.0 or 3.1 IMO.
One useful thing they could do would be to add 802.11ac wireless support. It probably won't come in their entry-level full-frame model, though, if it ever does.

[rawr]

but when I took photos of a raccoon's midnight raid I had to resort to manual focusing.

I think at some point, like your example, no matter what camera/lens you are using and how low-light sensitive the AF is meant to be, manual focus may be required.

Having said that, in your example, if you had a f2.8 or faster lens (with a good T-stop value and no filter) on your KP, rather than a slow zoom, I think the AF would probably have managed to work, and the ISO's would have stayed much more moderate too.

Ideally, for ambient low-light shooting a photog would have in-hand:

• low-light AF system +
• low-light capable metering sensor +
• fast bright lens +
• low-noise sensor +
• well balanced exposure and shutter settings +
• capable image processing software in-camera and on the PC to mash the noisy data into a good image

but in real life one has to make do.

[gatorguy]

I still do not see where they used Artificial Intelligence (A.I.) , except in the title of the article. They say that they trained the algorithm for the white balance, but then the algorithm is fixed, so by definition they aren't using A.I.

---------- Post added 16-02-19 at 18:45 ----------


Yes, you are right. Marketers now use AI everywhere even when they don't know what it is.

Here's the more complete explanation of how Pixel Night Sight works. There's a lot involved.
Google AI Blog: Night Sight: Seeing in the Dark on Pixel Phones

[biz-engineer]

Here's the more complete explanation of how Pixel Night Sight works. There's a lot involved.

The problem is, I'm done my PhD on Artificial Intelligence, and I don't see any of it used in camera today. To me, the articles claim that A.I is used but I don't see where it is used. Do users have to train their camera after unboxing?
Pixel phones use computational photography, and Pentax is one of the ILC brands using computational photography the most out of all brands. Pentax K1 II does pixel shift and hand held pixel shift, Canon and Nikon don't do it even with mirrorless.

---------- Post added 17-02-19 at 07:37 ----------

And one major part (or influence) has the photographer behind the camera. Mirrorless will be the future, but for the moment the "average at best" DSLR cameras can challenge the high end mirrorless cameras anytime. The secret is to stop watching cancan reviews and go out and shoot and gain confidence in yourself.

To my eyes, the Sony image looks better than the Canon image.. because they used a flash light with the Sony, and available light with the Canon, as a result, the colors are different and we see some shadows on the face on the image taken with Canon and no shadow on the face with the Sony. If a fash was used with the Canon, there would be a lot less difference between the two shots, if any.

[Dan Rentea]

To my eyes, the Sony image looks better than the Canon image.. because they used a flash light with the Sony, and available light with the Canon, as a result, the colors are different and we see some shadows on the face on the image taken with Canon and no shadow on the face with the Sony. If a fash was used with the Canon, there would be a lot less difference between the two shots, if any.

Yes, the Sony image looks better. It would be odd not to look better when you add quality light from flashes on the subject and on the ambient. But how much better is the Sony image given the fact that the Sony image was taken at ISO 1000 with flashes and the Canon image was taken at ISO 2500 with no flash? Not to mention that I shot at f2.8 with a cheap lens at he shot at f3.5 with a G master lens.

I'm telling you, if I had a flash, or if the Sony photographer would have shot with no flash at ISO 2500 instead of ISO 1000, you couldn't tell the difference in a blind test between the files.

And I have some other files taken with A9 and my 5D in the same day, but my wife doesn't want me to use the images with her as example on forum or social media.

[biz-engineer]

I'm telling you, if I had a flash, or if the Sony photographer would have shot with no flash at ISO 2500 instead of ISO 1000, you couldn't tell the difference in a blind test between the files.

Canon sensors have weaker DR (according to test charts) at low ISO, however, from ISO400 and up , there is no difference between sensors, they all produce about the same overall noise at equal sensor size. Anyway, cameras differences in the mind of people is largely inflated by internet forum rehearsal of the same things over and over again. The way you use the camera (lighting and framing) make a huge difference on the results, far more difference than what camera you use. Everyone should ask themselves if the new camera model will improve their photographs or if going to a camera class would do it better. For me the answer is clear, I'm not good enough for the gear I already have.

[Rondec]

Here's the more complete explanation of how Pixel Night Sight works. There's a lot involved.
Google AI Blog: Night Sight: Seeing in the Dark on Pixel Phones

But this stuff isn't AI it is computer enhancement of photos, which has been around for awhile, including on Pentax cameras. This seeing in the dark feature is basically the camera choosing a shutter speed it thinks the photographer can hand hold and then aligning and combining multiple images. I think Sony cameras have been doing this for quite awhile. Dynamic pixel shift gives this same sort of feature to the K-1 II.

Modern SLRs don't use as much computational enhancements of images, because higher end photographers really don't like them. They would rather do their own noise reduction after the fact, the same with things like skin smoothing and even stitching of panoramas. Still many modern ILCs have the ability to do these things if you do a bit of menu diving.

But AI stands for artificial intelligence which requires the computer to be interacting with the user in some way and learning and improving over time. However awesome the Google Pixel is, it isn't doing that.

[photoptimist]

But AI stands for artificial intelligence which requires the computer to be interacting with the user in some way and learning and improving over time. However awesome the Google Pixel is, it isn't doing that.

There are a lot of examples of AI (e.g., chatbots, autonomous vehicles, automated language translation, game playing) in which the "learning and improving over time" is only done in the lab by some engineers. That is, training the AI system is distinct from using the AI system.

There's also a much broader class of algorithms known as "machine learning" which help extract meaning and value from data of which AI algorithms such as "deep learning"(which are neural networks) are subset.

AI does NOT require interacting with a human user, it usually only involves the training algorithm working with training data. The major difference between AI and non-AI is the extent that the system relied on data to learn the patterns of what must be done rather than requiring a designer/engineer/programmer to explicitly program those patterns.

Thus Google Pixel may be an example of AI if Google used a data-based method such as deep learning to build the system.

[Rondec]

There are a lot of examples of AI (e.g., chatbots, autonomous vehicles, automated language translation, game playing) in which the "learning and improving over time" is only done in the lab by some engineers. That is, training the AI system is distinct from using the AI system.

There's also a much broader class of algorithms known as "machine learning" which help extract meaning and value from data of which AI algorithms such as "deep learning"(which are neural networks) are subset.

AI does NOT require interacting with a human user, it usually only involves the training algorithm working with training data. The major difference between AI and non-AI is the extent that the system relied on data to learn the patterns of what must be done rather than requiring a designer/engineer/programmer to explicitly program those patterns.

Thus Google Pixel may be an example of AI if Google used a data-based method such as deep learning to build the system.

I guess I read Google's explanation of their phone's combining of five images together to make a brighter image and it just doesn't sound different from techniques that have been around for ages. Maybe you are right and they've done something different, but it doesn't sound different to me. Same as Pentax uses auto alignment to enhance and combine HDR images or dynamic pixel shift.

[Dan Rentea]

Canon sensors have weaker DR (according to test charts) at low ISO, however, from ISO400 and up , there is no difference between sensors, they all produce about the same overall noise at equal sensor size.

Acording to those test charts, Sony A9 has 13.3EV when comes to dynamic range and Canon 5D Mark IV has 13.6EV. 5D Mark IV also has 30mp instead of 24mp so just by resizing the Canon images from 30mp to 24mp, the small advantage of Sony A9 at high ISO is gone. But Sony users have some cameras like A7R III with a dynamic range above 14.3EV and when they try to make a point they just say that Canon lags behind when comes to dynamic range. I don't disagree with this as long as Sony users specify also the cameras instead of speaking in general. That's why I came with examples and with arguments in response to Wider's comments regarding:

1. Canon and their average sensors at best - yet, his A9 which costs 1000$ more than 5D Mark IV has weaker dynamic range than the average Canon technology (by the way, the dual pixel af from the mediocre Canon technology is still unmatch by any mirrorless cameras)
2. A7 II has better af than Canon cameras; when I told him my experience with A7 II's af in low light he disagreed with me, but:

- the Sony manual seems to say that the af sensitivity of A7 II is -1EV instead of -3EV like Canon has on 6D/6D Mark II or -5EV for EOS RP
- the guy who posted his review after shooting with Sony for 4 years (the link posted by Winder) said the same thing I said when comes to the af in low light of Sony A7 II and I quote him "The A7 II’s autofocus struggled in low light, there was no autofocus point selector, and battery life was horrendous, especially in cold weather."

So, let's wait for EOS RP to become available and then we can talk again about how good the A7 II is compared to EOS RP. Because cancan reviews or test charts from tripod for me are equal to zero interest.

By the way, remove Nikon from the ecuation and see what J.P. Morgan has to say about EOS R vs Sony A7R III. And he is a Sony user and A7R III has 14.7EV dynamic range. Take the video with a grain of salt.

You Tube

You Tube

[Winder]

Acording to those test charts, Sony A9 has 13.3EV when comes to dynamic range and Canon 5D Mark IV has 13.6EV. 5D Mark IV also has 30mp instead of 24mp so just by resizing the Canon images from 30mp to 24mp, the small advantage of Sony A9 at high ISO is gone. But Sony users have some cameras like A7R III with a dynamic range above 14.3EV and when they try to make a point they just say that Canon lags behind when comes to dynamic range. I don't disagree with this as long as Sony users specify also the cameras instead of speaking in general. That's why I came with examples and with arguments in response to Wider's comments regarding:

What you do is cherry pick data points. Why are you comparing the A9 to the 5DIV? If we want to compare lets compare cameras that are actually going to target the same users. Compare the Sony A7III and the Canon 5DIV. The Sony A7III has a DR of 14.7 EV and costs under $2,000. The Canon 5DIV has an DR of 13.6 and a cost of $1,000 more than the A7III.

The A9 is designed for speed, AF, and to be used with the electronic shutter most of the time. The A9 competes with the 1DX which it matches or beats in pretty much every measurable category including price.


What about the A7rIII, which camera does that compare to? The 5DSR. How do those compare in terms of cost and performance.?


A7II would compare to the Canon 6DII or the new EOS-RIP.