Current Server Time: 15:25 (Central Europe)

#apertus IRC Channel Logs

2013/11/03

Timezone: UTC


23:41
FergusL
still around Bertl ?
00:17
Bertl
yup
00:42
FergusL
anything new today ?
00:43
Bertl
there is always new stuff, like the pll drp example :)
00:50
FergusL
haaa yes
00:50
FergusL
this mail looked liuke black magic to me
00:50
FergusL
though PLL rings a bell
00:50
FergusL
a distant one
00:50
FergusL
but I already heard this
00:52
FergusL
Bertl: can you write in query ? (just testing something with my client)
00:54
FergusL
looks ok Bertl, thanks
00:55
Bertl
np
01:08
FergusL
when coding
01:08
FergusL
I always struggle with trivial issues
01:08
FergusL
and spend lots of time just to enhance a few lines
01:08
FergusL
for the sake of elegance and speed
01:10
Bertl
nothing wrong with improving details
01:11
Bertl
the trick is in breaking problems down and testing every step
01:13
FergusL
http://scien.stanford.edu/pages/labsite/1999/psych221/projects/99/tingchen/algodep/vargra.html I'm trying to think about the best way to put the 5x5 region of interest into memory and access it to build the gradients
01:15
FergusL
yes and it doesn't help me, I'm not very fast
01:16
FergusL
so it's not even finished, nor remotely functional that I'm already polishing
01:52
Bertl
off to bed now ... have a good one everyone!
02:24
FergusL
night
02:28
troy_s_
joined the channel
02:29
troy_s_
left the channel
11:30
se6astian
joined the channel
11:30
se6astian
hello!
11:31
se6astian
http://www.redsharknews.com/technology/item/1208-first-images-from-axiom-alpha-prototype-camera-from-apertus
11:41
mars_
hi
11:53
se6astian
hello mars_
11:53
se6astian
just posted some thoughts about cooling the image sensor on the apertus forums: https://www.apertus.org/forums/viewtopic.php?f=4&t=1291&p=7369#p7369
11:54
se6astian
feedback appreciated
11:56
se6astian
some elphel news on their blog: http://blog.elphel.com/2013/11/nc393-development-progress-testing-the-hardware/
12:13
mars_
so do we have a peltier cooler? if not, we shoud order some and play around with it ;)
12:15
se6astian
well the prototype cant be equiped with one at this stage
12:16
se6astian
we would have to cut a hole into the PCB holding the image sensor
12:19
se6astian
we could just play around with a peltier element and a general source of heat, but currently its not so much of a priority
12:27
mars_
we can build bertl's proposed cooling scheme, and instead of the sensor we use other means of generating heat. and we can measure how much heat we can move out, what fans we need, ...
12:29
se6astian
yes thats what I meant with "general source of heat", our funds are very limited and this is an issue to be solved far in the future, so our priorities now are getting video from the HDMI port on the prototype and then raising money in crowd funding
12:36
mars_
i'm little confused now. what kind of feedback do you want on the cooling issue?
12:47
Bertl
morning everyone!
12:58
FergusL
morning as well from here !
13:38
se6astian
hello!
13:38
se6astian
feedback in the sense of ideas/concepts or agree/dissagree :)
14:22
FergusL
left the channel
14:23
FergusL
joined the channel
14:30
FergusL
left the channel
14:42
se6astian
Bertl, is a small 38mm diameter fan enough for the prototype you think?
14:43
se6astian
40mm square frame
14:49
FergusL
joined the channel
14:50
FergusL
left the channel
14:50
FergusL
joined the channel
14:59
Bertl
I guess it will be the only fan which fits
15:00
se6astian
on the sides yes, top bottom will be difficult as well
15:00
se6astian
I ll add a 40mm one to the design and we will see if that does the trick
15:01
Bertl
plan for a 40x40x20mm
15:01
se6astian
ok
15:01
Bertl
they are a lot more effective than the 40x40x10 or the 40x40x6
15:01
se6astian
ok
15:01
Bertl
and try to keep the space near the VGA port free
15:02
Bertl
i.e. put it on the other side if possible
15:02
Bertl
we might want to add a test power supply in the region near the VGA connector
15:14
se6astian
ok, I had the fan placed next to the VGA port, will move to other side
15:23
philippej
joined the channel
15:23
Bertl
hey philippej!
15:23
philippej
hello !
15:23
philippej
how is it going here ?
15:25
Bertl
fine, fine ... and on your end?
15:26
philippej
fine as well, enjoying the crapiest weather one can get in belgium, dirty rain, gray sky, cold wind :-)
15:28
Bertl
sounds not that much different from the weather we had when we visited you :)
15:29
philippej
exactly !
15:29
philippej
even worse in fact
15:30
philippej
I just received a fine for the travel to amsterdam, it was so rainy, you remember? I drove 10 km/h above the limit (110 instead of 100) : result : 40 euros :-(
15:30
Bertl
oww
15:33
FergusL
hi philippej !
15:33
philippej
that's life: -)
15:33
philippej
hello!
15:33
philippej
any news from the guy who said he was working on a raw -> dng converter?
15:34
Bertl
yeah, gabe is still working on the dng converter
15:34
se6astian
oh no, was it on the trip to or from amsterdam?
15:35
Bertl
haven't seen any code yet, but he promised me to send something soon
15:35
philippej
se6astian, to amsterdam, I'm the culprit :-)
15:36
FergusL
I'm working on a raw -> anything converter
15:39
FergusL
using OpenImageIO
15:41
FergusL
but I'm spending some time on the debayering part, working slowly
15:41
FergusL
http://www.pasteall.org/46990/python
15:45
philippej
if you have a dng fie, then you can play with all the debayering method available in libraw, so we could see what to expect from the sensor
15:46
FergusL
a dng file without debayering ?
15:47
philippej
dng is always without debayering afaik
15:47
FergusL
I'm implementing a rather good one, praised as the current best "public" one: http://scien.stanford.edu/pages/labsite/1999/psych221/projects/99/tingchen/algodep/vargra.html
15:47
philippej
ah well, then it's probably fine :-)
15:48
philippej
but still, having the dng headers added to the curent raw file would help
15:48
FergusL
let me see if I can do that
15:50
philippej
interesting read : http://www.libraw.org/about
15:54
FergusL
LibRaw has VNG (the method I'm implementing)
15:55
FergusL
iirc
15:55
FergusL
from dcraw
15:55
FergusL
it's also a personal trial at doing something cool :) and because I know python and have a hard time understanding the C code for debayering algorithm
15:57
philippej
of course
15:58
philippej
it's very interesting to test different algorythms
16:02
FergusL
Why so specifically DNGs ? Because it's what we have picked as a format ?
16:03
Bertl
obviously most commercial programms cannot read/process 'raw' data (and with 'raw' I mean really raw data)
16:04
FergusL
Of course, yes
16:06
philippej
it seems that dng is the only viable, somewhat long term, solution for storing raw files in a meaningful way
16:06
philippej
there is openraw, but with no rela momentum around it, I would not rely on it
16:06
philippej
(real)
16:08
philippej
source of raw2dng made for raw files produced by magiclantern (hacked canon cameras) : https://bitbucket.org/hudson/magic-lantern/src/tip/modules/lv_rec/raw2dng.c
16:09
FergusL
nice
16:09
philippej
there is a lot of room for improvement when converting pure raw to dng files
16:10
philippej
I'm sure all the manufacturers do a lot of tricks when writing their "raw" files
16:10
philippej
there is a huge room for work there, if you have some ideas, it would be great
16:11
FergusL
it's unclear to me what is to be expected from the dng file, how RAW it must be, how "tricked" from RAW it could be
16:12
philippej
in short : it should look good :-)
16:12
philippej
it means : no dead pixels, no dead rows or lines, corrected colors, etc...
16:12
philippej
I suspect the raw we get from cameras is much processed
16:13
philippej
curently the raw we get from the sensor is as raw as it can be
16:13
FergusL
for now I've gone for a different approach: taking the RAWest input, that is the .raw16 we get and make most the processing during the debayering, that is, pre- and post-debayered
16:13
FergusL
yes
16:13
philippej
Bertl, correct me if I'm wrong of course
16:14
FergusL
and I agree about the cameras, even the raw that they call RAW, hence my question about a "debayered" dng file
16:14
philippej
here, more interesting, a dng writer in C : https://bitbucket.org/hudson/magic-lantern/src/tip/src/chdk-dng.c
16:14
philippej
it's still bayer data, but processed
16:14
FergusL
yes
16:14
philippej
at least, black and white level corrected
16:14
philippej
in fact we have no ideas what others are doing
16:16
Bertl
for a start, we just want to have a format which can be processed by most 'professional' tools
16:16
Bertl
(for further investigation)
16:16
philippej
yep
16:16
FergusL
yes, RED cameras are black boxes, so is their RAW and their code
16:16
Bertl
I haven't had a problem to work with the .raw8 and .raw16
16:17
philippej
once it's dng it can be opened in any tool
16:17
FergusL
Bertl: debayered or not ? It's also a good idea to test existing debayering libraries
16:17
Bertl
raw is raw, debayering is post processing :)
16:18
FergusL
I'm not sure OIIO can write dng
16:19
philippej
I'd say it probably can't which seems logical, if it only works with debayered images?
16:31
philippej
See you all later, enjoy the end of the week !
16:32
philippej
left the channel
16:35
FergusL
troy_s: http://dannyruijters.nl/docs/cudaPrefilter3.pdf ?
17:12
Kush
joined the channel
17:12
Bertl
welcome Kush!
17:12
Kush
hello everyone, thanks :)
17:12
Kush
i'm FergusL friend
17:13
FergusL
hi Kush
17:14
FergusL
I'm having an audio chat for a few hours with Kush, discussing debayering and interpolation
17:17
se6astian
nice, what did you discuss in particular?
17:29
FergusL
well basically catching up with the concepts and looking into more complex algos as well
17:29
Bertl
the sky is the limit (for guessing data :)
17:31
FergusL
currently having a text chat with troy_s about cubic b w prefilter ( http://wiki.blender.org/index.php/User:Sobotka/Cubic_B-Spline_Scaling_Algorithm )
17:31
FergusL
and how to use it for debayering, which is his suggestion
17:37
se6astian
interesting
17:37
se6astian
what is currently the most widely/successful debayering algorithm you think=
17:37
se6astian
?
17:37
se6astian
AHD or VNG?
17:38
se6astian
Red claimed itterative debayering methods that require several passes are far supperior
17:38
se6astian
but I have never seen such an implementation that actually does mutliple runs (maybe its just marketing talk as all algorithmns do things in several steps....)
17:41
FergusL
I think it's AHD, but VNG (the one I'm working on) is great too, I'm not sure
17:41
FergusL
several steps, yes, several complete passes, I don't know
18:33
FergusL
nope, no .dng in oiiio
18:34
FergusL
there's .ppm
18:34
FergusL
(and a lot more obviously)
18:39
dmj_nova
yeah, oiio doesn't do dng
18:40
dmj_nova
but maybe it should be implemented
18:40
FergusL
it was asked for
18:40
FergusL
but it's understandable why it doesn't support it
18:41
dmj_nova
FergusL: you mean in that the developer doesn't have time or some other reason?
18:44
FergusL
no I mean that RAW is not what oiio is made for
18:45
dmj_nova
ah, makes sense
18:45
FergusL
and ultimately, using high precision float formats with one channel, you pretty much get your raw file
18:46
FergusL
but then on the other hand RAW software might not want to debayer from say an .exr
18:48
dmj_nova
yeah, you would need to debayer before OIIO
19:31
se6astian
left the channel
20:20
se6astian
joined the channel
20:20
se6astian
http://www.cinema5d.com/news/?p=21664
20:21
Bertl
interesting article :)
21:51
se6astian
ok time for bed
21:52
se6astian
if there is anything else I should bring to the happylab tomorrow send me an email
21:52
se6astian
as I will probably be on my way there before you are in your office yet :)
21:52
se6astian
good night
21:53
se6astian
left the channel