Google’s own camera app puts DNGs through a burst-stacking align-and-merge algorithm. I don’t think the HAL itself returns the stacked align-and-merge result, so I believe that third party apps will provide the results of a single shot.
That said, it could be that on Pixels, the HAL2 API actually returns an align-and-merge stack to anything that asks for a DNG… It’s kind of not clear.
There are definitely a lot of variables in phone-sourced DNGs, some of which cause piles of questions to be asked by people handling the data downstream.
In the past I’ve generally avoided dealing with this, but a recent vacation was to Disney World Resort and even an APS-C camera with a slow lens seemed like too much to lug around. (Even my mostly empty backpack didn’t always fit into the “please place your stuff here” cubby on some rides.)
Any Fujifilm X-T30 files required? I see that the camera support page at darktable.org says there are no WB Presets, no noise profile, and no custom matrix for this model.
Awesome, sample still wanted
Something like daylight landscape works best, as stated on RPU.
Err, that is not something that you have to do.
It’s the criteria i was looking for when saying that the sample
from that particular camera might be good to have.
I have a Ricoh Theta Z1 using the Dual Fisheye Plugin (which creates a stacked DNG, in my sample these are 7 exposures) and darktable doesn’t seem to be able to read the DNGs that it created. It WILL read the DNGs after I do some corrections in Lightroom…
I don’t really need darktable to be able to edit these, but figured I’d throw some samples to the community just in case.
Here is the out-of-the-camera file: R0020001_HDR.DNG (51.1 MB)
And the Dual Fisheye Plugin for the Theta Z1 camera:
It’s also worth noting that most (all except Lightroom???) of the image software I use won’t open that initial HDR DNG - GIMP, darktable, Windows Photo … NoMacs does seem to open it though. ¯_(ツ)_/¯
$ ./build-Clang-SANITIZE/src/utilities/rstest/rstest /tmp/R0020001_HDR.DNG -c
/tmp/R0020001_HDR.DNG : starting decoding ...
/tmp/R0020001_HDR.DNG failed: ../src/librawspeed/decoders/DngDecoder.cpp:386: virtual rawspeed::RawImage rawspeed::DngDecoder::decodeRawInternal(): Uncompressed float point must be 32 bits per sample.
Total decoding time: 0s
0x0102 Exif.Image.BitsPerSample Short 1 2 16 16
0x0153 Exif.Image.SampleFormat Short 1 2 3 IEEE floating point data
So it looks like to be an uncompressed 16-bit floating-point.
That’s an interesting choice, as if they were trying to use something that nothing supports. @Wayne_Schlegel Please file a bug?
Actually the DNG spec does not say compression is mandatory for floating point data, only “allowed”. So this seems to be a valid DNG, although unusual indeed…
In addition to the rawspeed bug report requested here, one for LibRaw is also needed please (confirmed it gives a similar pink output while the compressed one from LR is fine).
You could also ask your plugin developer to turn float DNG compression on so they get a wider reach in general.
@Wayne_Schlegel ping. @kmilos helped create the bugreport, but the fix still won’t be able to proceed without the sample.
Someone needs to contribute an appropriate sample to RPU,
under CC0 license.
I have uploaded a pair of phone pictures, that are useful e.g. for art dng embedded data correction. poco m3 few days ago and p20 lite in october, but i have seen they haven’t been uploaded.
Keep in mind that they’re operating on a device with severely limited CPU here, to the point where compression often throttles save-out rates significantly worse than flash memory write speed limitations of uncompressed data.
I’ve rented Nikon d780 for the weekend, do you need sample raw files (NEF) from this camera? They are not on the list of supported cameras on darktable.org website.
Just notice that only extra data is missing and raw files for this camera is there. Can you do something to support those other things in the table? WB Presets, Noise Profile, Custom Matrix.