I just updated to 18.104.22.168 on my macStudio (Ventura). I was processing a batch of OM-1 raw files from a nighttime parade last night. When I examined the DPXD output (jpeg 95%) the DPXD images were worse than the original raws. I went back and ran again using DP, and these images are what I expected: much improved noise, particularly for faces. But the DPXD seems to have failed completely. I had tested DPXD before on other images and it worked fine (and slightly better than DP) so this is a regression. I wonder if anyone else has seen this - it’s pretty obvious, but might be associated only with the OM-1 raws.
There is an issue with DPXD and Ventura. Images all have a purple cast. There is a long thread on this, plus a workaround someone found until DxO come up with a patch.
It’s not a color cast - it just doesn’t remove the noise (at all). I guess I need to file a support request.
It’s not uncommon to find (in an image-specific sense) that DP-XD is not providing much benefit over DP - but I am surprised to hear that it’s worse.
There’s discussion here on the type of image for which DPXD is best suited - with general consensus that it’s worth applying mostly on heavily cropped images … Otherwise, there’s often no visible advantage over DP (and, DPXD comes with a processing time cost).
- So, perhaps the simplest approach for you is to use DeepPRIME (and not DeepPRIME XD).
Which lens(es) are you using with your OM-1 ?
It’s not that DPXD isn’t as good as DP - it’s that it seems to have no effect at all! The processed images look just like the raw image without any noise reduction applied at all. I have 110 images shot with the OM-1 and the Oly f2.8 12-40 Pro lens and I see that on every image. I have tried a few tests with the 40-150 f2.8 Pro lens and those seem OK - the same effectiveness as DP. I also tested a while ago with a variety of cameras and lenses and DPXD worked fine - but very slow on the 61mPix Sony A7RM4 images. I will do some additional testing tomorrow.
Update. I have processed more OM-1 images (on my MBP16M1ProMax) and don’t see the issue. DPXD is definitely not better than DP for these images (in fact, it is a little worse) but I need to do more testing to understand what is happening. At this point I don’t see any advantage here worth the 3x processing time.
Tested again on iMac running macOS Monterey. DeepPRIME and DPXD work here and produce different results, e.g. in a high iso night take of foliage in a backyard.
Whether I prefer one processed image over the other is a matter of personal preference though.
Same test on M1 MBA with Ventura delivers the already mentioned colour shift, but the difference is quite noticeable:
I was shooting with the OM-1 at ISO 3200 at f2.8. (4/3 cameras are probably not the best choice for low-light shooting.). I was looking specifically at faces - and pretty consistently, the DPXD results looked a little worse. I can see it it trying harder to bring back a little detail - but it isn’t an improvement for faces in my opinion. I am running Ventura and I do see the slight purple tint issue.
In the past, I have seen very good results with my Sony A7RM4. I am waiting for my order of the A7RM5 to see how it works there - probably mostly the same, given that the M4 and M5 have the same sensor (though the camera processing is different for the M5).