Current Server Time: 16:39 (Central Europe)

#apertus IRC Channel Logs

2018/06/10

Timezone: UTC


03:22
rton
left the channel
03:44
Bertl_oO
off to bed now ...
03:44
Bertl_oO
changed nick to: Bertl_zZ
05:51
futarisIRCcloud
joined the channel
06:45
ymc98_1
joined the channel
06:47
ymc98
left the channel
08:53
BAndiT1983|away
changed nick to: BAndiT1983
09:15
niemand
joined the channel
11:19
ymc98
joined the channel
11:22
ymc98_1
left the channel
12:13
Bertl_zZ
changed nick to: Bertl
12:13
Bertl
morning folks!
12:15
ymc98
left the channel
12:56
ymc98
joined the channel
13:03
ymc98
left the channel
13:06
ymc98
joined the channel
13:21
ymc98
left the channel
13:29
rton
joined the channel
13:47
aombk2
left the channel
14:36
illwieckz
left the channel
14:48
illwieckz
joined the channel
15:24
ymc98
joined the channel
15:31
aombk
joined the channel
16:07
RexOrCine|away
changed nick to: RexOrCine
17:00
ymc98_1
joined the channel
17:03
ymc98
left the channel
17:59
se6astian|away
changed nick to: se6astian
18:12
illwieckz
left the channel
18:13
illwieckz
joined the channel
18:20
illwieckz
left the channel
18:24
felix_
i'll be at the defcon this year; i don't know that much details about the project apart from the sdi subproject i'm working on though. if there's some sticker exchange box and i get some more stickers, i can put them there; i think i already distributed most of the stickers i got
18:24
illwieckz
joined the channel
18:26
illwieckz
left the channel
18:29
Kjetil
Don't bring your laptop ? :P
18:31
illwieckz
joined the channel
18:57
illwieckz
left the channel
19:05
felix_
i'll probably only bring my old laptop ;)
19:07
felix_
new one isn't ordered yet though, since i want a configuration that you can't configure in the online ordering system; so i'll have to order by telephone...
19:07
illwieckz
joined the channel
19:12
TofuLynx
joined the channel
19:12
TofuLynx
Hey BAndiT1983
19:12
TofuLynx
we can talk here :)
19:13
RexOrCine
felix_ You want more information on the project?
19:13
felix_
sure
19:14
RexOrCine
You seen the brochure?
19:14
felix_
not sure
19:14
RexOrCine
Have a look at documentation on the right hand side - https://wiki.apertus.org/index.php/AXIOM_Beta
19:15
RexOrCine
Also Project Background on the left hand side.
19:15
felix_
haven't seen the brochure before
19:17
felix_
i've put the brochure and documentation on my reading list
19:19
illwieckz
left the channel
19:19
illwieckz_
joined the channel
19:20
illwieckz_
changed nick to: illwieckz
19:24
BAndiT1983
hi TofuLynx
19:24
TofuLynx
So, I will fix the unit tests and the SHOODAK Debayer image borders
19:25
TofuLynx
Can you explain the fuzziness of SHOODAK? xD
19:28
se6astian
define "fuzziness"?
19:30
TofuLynx
He is running the SHOODAK debayer I made and he commented that the result is fuzzy a lot, compared with GEDI Debayer and Bilinear Interpolation :)
19:33
se6astian
please share links to images
19:35
BAndiT1983
it has a lot of uneven edges, just fuzzy, don't know how to explain it better
19:36
TofuLynx
Okk! Makes sense, due to the randomness associated with it
19:36
TofuLynx
The fuzziness can bring problems in video, as it will be always changing the random offsets
19:37
se6astian
well shoodak is meant to be a lot softer than most debayering methods nowadays, its a feature not a bug :)
19:37
TofuLynx
But yeah, each algorithm has its pros and cons
19:37
se6astian
but compared to bilinear there should not be that much difference
19:38
TofuLynx
SHOODAK2 on paper is roughly comparable to nearest neighbor interpolation, with a difference in the green channel, which is random nearest neighbor
19:38
BAndiT1983
looking forward to see it in motion
19:38
TofuLynx
and that results in very sharp images, but with a lot of artifacts in edges
19:39
TofuLynx
I implemented a edge sensing algorithm, but it is very simple and doesnt do wonders right now
19:39
TofuLynx
Yeah, me too
19:42
BAndiT1983
TofuLynx: have you tried the new switch?
19:42
TofuLynx
Not yet
19:42
BAndiT1983
ok, you can see it a lot better when doing live switching
19:43
BAndiT1983
have used our test dng with Bertl soldering
19:43
TofuLynx
hmm
19:43
BAndiT1983
if oyu zoom to the glass, then you can see it better
19:43
TofuLynx
Let me guess, there are a lot artifacts in the bird pencil?
19:43
BAndiT1983
*you
19:43
TofuLynx
Ok!
19:43
TofuLynx
The glass is a very trick zone!
19:43
BAndiT1983
just zoomed to glass, as i knew, that it's problematic for debayering
19:44
BAndiT1983
also haven't looked into the pipeline yet, to see when defringing has to be done
19:44
TofuLynx
how does the switch works? does it render each time it switches, or it renders every debayer at start?
19:44
BAndiT1983
defringing is when you adjust offset of blue and red channels
19:45
BAndiT1983
it renders always, just stops when "file open dialog" is coming up, to prevent problems, then it reloads the image with selected debayer method
19:45
TofuLynx
What are the main reasons to defringe?
19:45
BAndiT1983
no tricks yet, as i wanted to get it done quickly
19:45
BAndiT1983
more clear image
19:45
BAndiT1983
http://img.linuxfr.org/img/687474703a2f2f7777772d766572696d61672e696d61672e66722f7e6d6f792f746d702f64656672696e67652e706e67/defringe.png
19:46
BAndiT1983
colors are more correct
19:46
BAndiT1983
https://www.youtube.com/watch?v=cruud_MEr_U
19:47
TofuLynx
I see
19:47
TofuLynx
where do you notice it more?
19:47
BAndiT1983
don't remember at the moment, but there were some places
19:47
BAndiT1983
i think regions with higher contrast suffer from it more
19:48
BAndiT1983
we can orintate on darktable or rawtherapee pipeline, to be sure
19:48
BAndiT1983
*orientate
19:48
TofuLynx
Yeah, I see
19:55
TofuLynx
I had no idea that there were a lot of processes behind image correction and debayering
20:09
TofuLynx
left the channel
20:09
TofuLynx
joined the channel
20:25
g3gg0
joined the channel
20:25
g3gg0
hi
20:26
BAndiT1983
hi
20:27
g3gg0
did someone calculate the PLR kneepoints from Vtfl2/3 and Exp_kp1/kp2 already?
20:29
BAndiT1983
nope
20:29
BAndiT1983
is it important at the moment?
20:30
g3gg0
depends on your definition of important :)
20:31
BAndiT1983
in my opinion, it's not that important, as infrastructure should be there, before doing adjustments etc.
20:33
g3gg0
i partially agree. we should know how to do things by hand before we start proper implementations
20:34
g3gg0
first prototype and know how it works and what the quirks are, then sketch an implementation
20:35
g3gg0
in this case i wonder if its better to convert the raw multi slope related register values to coordinates which describe how to map the raw payload
20:37
BAndiT1983
i would go with array for now as LUT and consider more complicated methods for later
20:37
g3gg0
where can i find what vhigh is set to?
20:37
g3gg0
where do you get your LUT from?
20:39
BAndiT1983
it should be coming from calibration maybe
20:39
BAndiT1983
and stored somewhere
20:40
g3gg0
uhm kind of calib () { <magic happens here> }
20:40
g3gg0
calibration is for finding differences to what value is being expected
20:40
BAndiT1983
https://wiki.apertus.org/index.php/Sensor_Calibration
20:41
g3gg0
so you know your curves and algorithms and calibration is for finding offsets like dark currents etc
20:41
BAndiT1983
i'm not that deep in that stuff, it was done by other guys
20:42
g3gg0
i know. looks like alex worked out the PLR page from how it is structured
20:42
BAndiT1983
thought so
20:42
g3gg0
@alexML you there?
20:46
BAndiT1983
hm, that color chart reminds me of color correction in resolve, where you can drag a set of colored rectangles and span them on the test pattern
20:46
BAndiT1983
maybe it can be automated with opencv
20:48
g3gg0
thats its purpose ;)
20:48
g3gg0
iirc resolve only supports x-rite and one/two others
20:49
BAndiT1983
i know the purpose, was just thinking about automating the process, just an idea for the future maybe
20:50
g3gg0
nah thats the wrong way. first thing is to totally understand how it works and how its being calculated
20:51
g3gg0
in the field, you will always have offsets, drifts and stuff you cannot predict. thats what calibration is for
20:54
BAndiT1983
i know, my thoughts were just going in another direction, as i'm looking for ways to make the life of users simpler
20:54
BAndiT1983
but we have a long road till then
21:05
TofuLynx
BAndiT1983, do I need to submit anything to Google for the first evaluation? Or is it just answering the questions?
21:06
BAndiT1983
i think just questions
21:06
BAndiT1983
but when you are ready to merge, then we can also do this step, so the repos are synchronized
21:06
TofuLynx
Ok! :)
21:07
TofuLynx
Hmm, to merge what? The unit tests?
21:08
BAndiT1983
everything from your repo, new debayer stuff etc.
21:08
BAndiT1983
looked at debayering methods with new processingtest features
21:09
TofuLynx
ah! into the original fork?
21:09
BAndiT1983
shoodak is really fuzzy
21:09
BAndiT1983
yep, into original one
21:09
TofuLynx
Ok! I think we can merge after I create the new unit tests and fix the SHOODAK2 borders
21:09
BAndiT1983
from your dev to original dev, then build, afterwards i can merge dev to master
21:09
BAndiT1983
alright
21:10
TofuLynx
By the way, Bayer PreProcessor crashes sometimes with segfault, do you know this issue? It's an old issue I think
21:10
BAndiT1983
there is a keyboard in the test image where SHOODAK also shows the problems
21:10
BAndiT1983
maybe some race condition, do you know where it fails
21:10
BAndiT1983
*?
21:11
TofuLynx
No I don't, didn't look further into it
21:13
BAndiT1983
just copy the stacktrace if it fails again, there should be the line of crash
21:14
BAndiT1983
there is also a crash for some other DNG files than apertus ones
21:14
BAndiT1983
will investigate it
21:16
g3gg0
left the channel
21:22
TofuLynx
Roger!
21:31
BAndiT1983
very funny, image that crashes is DNG, but has just 8bit data
21:35
niemand
left the channel
21:41
BAndiT1983
TofuLynx: will push adjustments for bits per pixel recognition, was set to fixed value till now
21:41
TofuLynx
what is the issue with bits per pixel?
21:42
BAndiT1983
it was set to 12 all the time, but now the value of tag 258 is used correctly
21:43
BAndiT1983
added integer8 also, but just to prevent crashes with such images, output is black, will investigate further, at least it doesn'T crash right away, but you cannot open another image as the mechanism is not fixed yet, then it crashes, as the buffer is not allocated again
21:44
BAndiT1983
memory pool and caching have to be implemented soon
21:44
TofuLynx
Okk!
21:45
TofuLynx
I will probably push my changes later this night, so you probably will see the new commits tomorrow
21:45
BAndiT1983
ok
21:45
TofuLynx
Do you want me to do a Pull Request immediately after it?
21:45
TofuLynx
Or only after you build my commits?
21:45
BAndiT1983
if you think, that you are ready, then feel free
21:46
TofuLynx
Sure! :)
21:46
BAndiT1983
your commits are built by travis, i rely on it
21:46
BAndiT1983
https://travis-ci.org/TofuLynx/opencine/builds
21:46
TofuLynx
ok!
21:46
TofuLynx
those fails due to unit tests... xD
21:46
BAndiT1983
and just now came the email with "still failing" after my push ;) unit tests are broken
21:47
BAndiT1983
my goal was to separate build and unit tests in travis, but it seems a bit more complicated
21:50
BAndiT1983
ah, found something for supragya, maybe he already tried it
21:50
BAndiT1983
there is extract option in exiftool, also for linearization table
21:51
BAndiT1983
CFA Repeat Pattern Dim : 2 2
21:51
BAndiT1983
CFA Pattern 2 : 0 1 1 2
21:51
BAndiT1983
Linearization Table : (Binary data 1244 bytes, use -b option to extract)
21:55
TofuLynx
I remember that when I used exiftool
21:57
TofuLynx
what is his current task?
21:59
BAndiT1983
have to look into logs, as there are so many things going on when discussing his tasks, so i'm losing track of it, we need a better roadmap, otherwise it's drifting somehow off
22:00
BAndiT1983
i think that currently DNG conversion is his goal through MLV stuff
22:01
BAndiT1983
still not sure about pefromance by using FUSE, but will wait for results
22:02
BAndiT1983
TofuLynx: should we add "none" as debayer methods for debugging purposes, so one can see the image before processing?
22:02
BAndiT1983
*method
22:03
TofuLynx
Hmm
22:03
TofuLynx
that's a good idea, shouldn't be hard too
22:04
BAndiT1983
it's not hard, will add it, still struggling with the test image, as it gives black, even without debayering, so the problem is in extraction
22:05
BAndiT1983
will also add the dump feature you asked for, but a bit later
22:05
se6astian
changed nick to: se6astian|away
22:05
TofuLynx
Sure! :)
22:11
BAndiT1983
done and pushed
22:15
TofuLynx
Ok! Thank you! Loved the new interface!
22:15
TofuLynx
Do you have experience creating UIs from the past or current work?
22:17
BAndiT1983
both
22:17
BAndiT1983
and learned something funny about qt
22:18
TofuLynx
Nice :P
22:18
BAndiT1983
if you want to shrink panels down to the content, you have to set it to "fixed"
22:18
TofuLynx
it doesnt even make sense
22:18
BAndiT1983
can't believe how dumb this is, as there is "minimum" but it's not working in that way
22:18
TofuLynx
xD
22:18
TofuLynx
should be like "auto"
22:19
TofuLynx
or like HTML
22:21
TofuLynx
I have to go
22:21
TofuLynx
See you next time! :)
22:21
TofuLynx
left the channel
22:21
BAndiT1983
see you, good night
22:31
BAndiT1983
also off, good night
22:31
BAndiT1983
changed nick to: BAndiT1983|away
23:18
comradekingu
left the channel
00:12
jucar
left the channel