cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

Canon Neural Network Image Processing Tool - Software New Technology Release

shadowsports
Legend
Legend

Canon Neural Network Image Processing Software was released 11/21/2022 for the R5 C, and other MILC bodies. It integrates with DPP. This is a subscription-based product. More below if you want to read about it.

Canon : Product Manual : Digital Photo Professional : Neural network Image Processing Tool (start.ca...

image.canon | Cloud RAW image processing

~Rick
Bay Area - CA


~R5 C (1.0.6.1) ~RF Trinity, ~RF 100 Macro, ~RF 100~400, ~RF 100~500, +RF 1.4x TC, +Canon Control Ring, BG-R10

~DxO PhotoLab Elite ~DaVinci Resolve ~Windows11 Pro ~ImageClass MF644Cdw/MF656Cdw ~Pixel 8 ~CarePaks Are Worth It
18 REPLIES 18

jrhoffman75
Legend
Legend

Thanks Rick. Interesting product. A little more info here:

A sophisticated image processing tool using deep learning technology

Says it supports the 1D X Mark III, yet also provides a list of lenses it is restricted to and they are all RF lenses.

John Hoffman
Conway, NH

1D X Mark III, Many lenses, Pixma PRO-100, Pixma TR8620a, LR Classic

jrhoffman75
Legend
Legend

I tried the free one-month subscription. (Need to register with a credit card, but can go right back and cancel the auto-renew.)

Before/after at 200% on a random image I selected. (1D X Mark III, ISO 6400.)

DLO in the Neural Network is limited to certain RF lenses; a notice pops up to that effect.

Original

Processed

John Hoffman
Conway, NH

1D X Mark III, Many lenses, Pixma PRO-100, Pixma TR8620a, LR Classic

can't see the images

wq9nsc
Authority
Authority

Thanks Rick and I will probably give it a try. I am picking up a second 1DX III body so it will be interesting for football and other night sports. I am not impressed that they won't be providing support for any of their classic EF "great white" primes but I haven't been impressed with much that Canon has done during/post-pandemic so color me not surprised.

Rodger

EOS 1DX M3, 1DX M2, 1DX, 5DS R, M6 Mark II, 1D M2, EOS 650 (film), many lenses, XF400 video

I loaded it up tonight and for some reason, it wouldn't activate. I finally, out of frustration, restarted my system (Win 10 Pro), and it gave me the -Start The Tool- menu when I cranked up DPP. I bought the year with the intention of trying it at least a month, but I think I'll keep it and see if they improve or add to it.

I have one of the lenses it supports, the RF 15-35mm f/2.8L USM, and have some shots that I took of the moon behind clouds and Jupiter that are OK full frame, but noisy when zoomed in. Granted, these are extreme zooms that I doubt anyone would use, but it was a good test for the software. I also tried it on an RF 100-500mm L w/1.4X extender. As expected, it noted in the panel that it would not perform lens correction, but it did a really good job of cleaning up the image. These were shot with the EOS R5, hand held. That white dot, down and to the right of the moon/clouds, is Jupiter.

BTW, the CRN files this program creates are 4X bigger than the CR3's they were made from.

RF 15-35mm f/2.8L USM, f/2.8, ISO 8000, 1/40th. My intent was to bring the trees out a little better because they were fairly well lit, considering 🙂RF 15-35mm f/2.8L Full Frame CRN.RF 15-35mm f/2.8L Full Frame CRN.RF 15-35mm f/2.8L CR3 Before Neural Net at 200% Zoom.RF 15-35mm f/2.8L CR3 Before Neural Net at 200% Zoom.RF 15-35mm f/2.8L After Neural Net at 200% Zoom (CRN).RF 15-35mm f/2.8L After Neural Net at 200% Zoom (CRN).

RF 100-500mm L w/RF 1.4X Extender, ISO 400, 1/800th, f/10, 700mm.RF 100-500mm L w/1.4X Ext. CR3 Before Neural Net at 100% Zoom.RF 100-500mm L w/1.4X Ext. CR3 Before Neural Net at 100% Zoom.RF 100-500mm L w/1.4X Ext. After Neural Net at 100% Zoom (CRN).RF 100-500mm L w/1.4X Ext. After Neural Net at 100% Zoom (CRN).

Newton

EOS R5, R6, R6II. RF 15-35 f/2.8L, 50mm f/1.2L, 85mm f/1.2L, 100mm f/2.8L Macro, 100-400mm, 100-500mm L, 1.4X.

This seems to be a monetized version of Cloud Computing that was in DPP a year or so ago.

It seems to work on "pure" RAW files. I selected DLO in DPP for a non-supported lens and got the message that DLO was not performed. Selected Monochrome Picture Style in DPP and a color image was returned. Similarly a deliberately underexposed file was exposure adjusted in DPP and an underexposed file was returned.

From a noise treatment standpoint I achieved similar results with Topaz Photo AI.

I think the real benefit will be using the tool with supported lenses so DLO can be applied.

John Hoffman
Conway, NH

1D X Mark III, Many lenses, Pixma PRO-100, Pixma TR8620a, LR Classic

I agree, John. After reading your post, I ran the files through Topaz and it does just as good a job. Although Topaz is more expensive, I already have it and you have more control compared to the two simple sliders with Neural Net (actually just one for unsupported glass). Yes, you can continue to edit CRN files in DPP, but without DLO, it's just not worth it, IMO. Also, existing DLO in DPP works fine for the one supported lens that I have.

I canceled my subscription. Now I just have to wait and see if Canon gives me a refund because when going through the cancelation process, I got several conflicting messages about canceling. It clearly states, when you are subscribing, that you have 31 days with no charge/full refund. But when you cancel, it's a different story. I guess we will see 🙂

Newton

EOS R5, R6, R6II. RF 15-35 f/2.8L, 50mm f/1.2L, 85mm f/1.2L, 100mm f/2.8L Macro, 100-400mm, 100-500mm L, 1.4X.

John,

When I tried the cloud computing noise reduction for DPP when it was embedded in DPP, it didn't do any better work than the algorithm built into DPP. My thought at the time was it would be useful for people with fast internet but not a lot of processing power. The dual CPU HP workstation I use has no problem chewing on 1DX III files so after trying a few medium and high ISO night sports photos on both the cloud and internal DPP options and finding no difference, I quit bothering with the cloud box.

As soon as I have time, I will give the Canon subscription service a test and see how it does compared to Topaz on some difficult captures. I like Topaz OK but they are as bad as Corel video studio in constantly nagging trying to sell upgrades and additional products and I wouldn't mind saying goodbye to them if the Canon product compares well even though Canon is no longer in my top 5 list of favorite companies 🙂

Rodger

EOS 1DX M3, 1DX M2, 1DX, 5DS R, M6 Mark II, 1D M2, EOS 650 (film), many lenses, XF400 video

wq9nsc
Authority
Authority

I purposefully shot some high ISO files last night at a girls' basketball game to give the neural networking tool a little test and I am fairly impressed with the results.  And I like that it makes good use of the Nvidia GPU "Cuda cores" unlike DPP basic which has gotten better, the software is pretty fast on my dual CPU/dual Nvidia workstation graphics card system.

What I don't like:

1.  It needs a keyboard shortcut to invoke the process, Canon does an excellent job of human factors interface with their 1 series bodies but their software frankly sucks in this important area.  And don't prevent it from being invoked when in the crop window, just another useless slowdown that shows very poor attention to human  factors.  I am glad that the Canon software team doesn't design cars or aircraft.

2. Give us an option to auto close the stupid nag box once processing is completed.  DPP updates always default to having to check the option box to close the image processed nag box but I don't even see the option box with the neural network tool.

Like some of the mirrorless firmware, Canon seems to have gotten into the habit of releasing stuff not quite ready for prime time.  I like what it does but maybe it is time to get a little user feedback from beta to catch this type of shortcoming.

Rodger

EOS 1DX M3, 1DX M2, 1DX, 5DS R, M6 Mark II, 1D M2, EOS 650 (film), many lenses, XF400 video
Announcements