00:00 | intrac | left the channel | |
01:04 | intrac | joined the channel | |
01:49 | illwieckz | left the channel | |
06:18 | illwieckz | joined the channel | |
10:00 | se6astian | anuejn: got a reply from polyrhythm, he is using https://www.rawdigger.com/ to open the DNGs
| |
10:24 | se6astian | so I now compared how different tools handle the DNG
| |
10:24 | se6astian | here is the overview
| |
10:24 | se6astian | https://cloud.apertus.org/index.php/s/jFcXQAfEem5TTnb?path=%2Fanejns%20converted%20dngs
| |
10:25 | se6astian | Adobe Photoshop - opens but is completely white/grey - no actual image content
| |
10:25 | se6astian | Adobe Lightroom - same as Photoshop both use Adobe Camera Raw Importer
| |
10:25 | se6astian | RawDigger - hints of image elements visible rest looks noise/error/random
| |
10:25 | se6astian | Rawtherapee - looks the same as in adobe camera raw, just white/grey
| |
10:25 | se6astian | Darktable - yes
| |
10:25 | se6astian | screenshots of each tool at link above
| |
10:25 | se6astian | in summary darktable is the only tool that displays an actual image that is expected
| |
10:27 | BAndiT1983 | nice summary!
| |
10:31 | se6astian | anuejn: I would check the metadata fields for things that might be required to interpret the raw data properly
| |
10:31 | se6astian | you can compare to the raw2dng converted axiom DNGs
| |
10:31 | se6astian | http://files.apertus.org/AXIOM-Beta/snapshots/sample-images/
| |
10:31 | se6astian | charts-x1-20ms.DNG
| |
10:31 | se6astian | portrait-gainx2-offset2047-20ms-01.DNG
| |
10:31 | se6astian | both open fine in rawdigger
| |
10:32 | se6astian | whitelevel and blacklevel might be values that are essentially required
| |
11:37 | anuejn | oh interesting :)
| |
11:38 | anuejn | yeah probably only darktable is actually implementing floating point dng
| |
11:38 | anuejn | I will look into it :)
| |
11:55 | anuejn | okay so it seems that the problem is not really missing tags, since it also does not work when using the adobe dng converter before
| |
11:55 | anuejn | looks like I have to build a node for converting floating point data to uint16
| |
12:05 | anuejn | also interesting software with an interesting blog https://www.fastrawviewer.com/blog/determining-practical-dynamic-range
| |
12:05 | anuejn | sounds like a cool experiment to do with the beta
| |
12:10 | BAndiT1983 | is the data originally received as floating point from the sensor?
| |
12:15 | se6astian | BAndiT1983: no, 12bit raw
| |
12:15 | se6astian | anuejn: great that we have a way forward now!
| |
12:16 | se6astian | will read blog post
| |
12:16 | BAndiT1983 | then still wondering about the floating and uint16 conversion
| |
12:22 | anuejn | BAndiT1983: we are averaging a bunch of frames together to reduce the dynamic noise
| |
12:23 | anuejn | that is done in floating point
| |
12:39 | se6astian | anuejn: you saw https://www.rawdigger.com/node/268 I assume?
| |
12:40 | se6astian | "floating-point DNGs are currently not supported"
| |
12:40 | anuejn | ah no i didnt see that
| |
12:40 | anuejn | but makes perfect sense
| |
12:46 | se6astian | I just also tried in Davinci Resolve - does not throw an error when loading the file but displays it as "media offline" with wrong metadata 2048x1556 resolution for example
| |
12:46 | se6astian | added to txt file on nextcloud
| |
12:48 | se6astian | https://paste.niemo.de/orozupumul.yaml
| |
12:49 | se6astian | output from https://github.com/antonwolf/dng_stacker/blob/master/dng_validate.exe
| |
12:50 | anuejn | seems actually fine?
| |
12:50 | anuejn | also tried davinci for the non-float images and it worked
| |
12:50 | anuejn | seems like float support is just very poor generally
| |
12:53 | se6astian | yes, only warnings is good I would say
| |
12:53 | se6astian | yes float seems uncommon for cameras
| |
13:04 | se6astian | read the article, I would be interested in what he means by "resolution is evaluated using a semi-automatic computerized method"
| |
13:40 | se6astian | if we know that it would indeed be interesting to also try that method with the beta
| |
15:15 | anuejn | I have to admit that semi-automatic computerized method does not sound very scintific
| |
15:15 | anuejn | but yeah that is a very interesting plot
| |
15:15 | anuejn | we could also try to produce such plots for different other sensors we own and then compare
| |
17:16 | intrac | left the channel | |
17:17 | intrac | joined the channel | |
17:54 | anuejn | okay so I build the FP32ToUint16 Node and now I can open the DNG files in rawdigger :)
| |
17:55 | anuejn | there is some documentation here: https://github.com/apertus-open-source-cinema/axiom-recorder/blob/24c0958d1d1dee7b910ac6237a089003be116920/README.md#dng-output
| |
17:57 | anuejn | I also converted the files to dng with uint16
| |
17:57 | anuejn | the archive is currently uploading
| |
17:57 | anuejn | I converted the files with a 16x multiplier to rescale the native 12bit output of the sensor to the 16bit dng
| |
18:05 | anuejn | here are the files: https://files.niemo.de/221201_XriteSnaps_uint16_dng.zip
| |
18:08 | anuejn | (polyrhythm, se6astian)
| |
23:24 | Guest58 | joined the channel | |
23:25 | Guest58 | left the channel |