#apertus IRC Channel Logs

morning folks!

Hi se6astian Bertl_oO : Have the znc servers been set up successfully?

hi supragyaraj

rather okay, and you?

I will send you and g3gg0 an email of my emulation program design

ok, looking forward to it

I will send an email... the explanation would be lengthy

I will have the stream saved in memory and then given off to the "recorder" as a stream

this stream will maybe be served through fuse or something.. and then we flush the saved MLV to memory itself.

and there are no bottlenecks...

FUSE would be overkill, in my opinion

https://github.com/supragya/AXIOM_RawStreamHandler/tree/master/src

generator: creates MLV vidf structs etc (which in real life would be given off by AXIOM)

stream_handler: Our actual system

stream_handler is multithreaded for each source/stream connected.. (more on it in email)

Hi, I'm a newbie and would like to start contributing as a software developer. Please advise about where I can get started from

how are you manan1999_

in terms of software development? your skills?

but I'm open to and looking forward to learning something new

I looked at it but would like a further insight

and then there is website itself (however I am not sure how much of it's access is given to public), it wont be on flask however

I'm afraid I don't have experience with any of them

there is also PCB inspection, using opencv through python, frontend done with nodejs

or newsletter templating

ah yes, BAndiT1983 that!

what happened to it? is it not working

how does newsletter templating work

you said last newsletter was made manually

I guess you were upto it? we have a lab task for it?

idea was to use templating library to bypass that and to insert texts and images automatically, from some content file

the templating was done a little by me... and later taken over by BAndiT1983 who did quite some work on it.

manan1999_: If you would like to work on it, I would help you begin on that task... it should be a good fit for you.. if you like

when can I start

free right now ?

yeah

yup

Hello BAndiT1983

Hello raj!

bogged with some things?

hi TofuLynx

I am seeing your bilinear implementation right now

could you write a test for it, so we can compare output with newer implementation?

what do you mean?

yep

ok

https://github.com/TofuLynx/opencine/blob/777ee30e8cdf6d6a9ba9ee6b8152f3034e8c8dae/Source/OCcore/Image/BilinearDebayer.h

Yeah

let's assume we are handling a RGGB image

so the algorithm know where to start

why isnt Red 0 and 1 then?

what's a pixel for you? the 4 colors?

rggb are 4 pixels

well, actually we call them 'sensel'

Sure, lets go with sensel

it makes it simpler to discuss pixel (RGB, HSV, CMYK, etc) vs photo sites (sensel)

Hmm what about the green channels, why is it 1 1 and 2 2?

https://www.researchgate.net/profile/Emanuel_Marom/publication/45709080/figure/fig1/AS:307400574816256@1450301416028/Typical-Bayer-Matrix-with-3-main-colors-in-the-RGGB-configuration-In-this-paper-each.png

hey supragya and TofuLynx, what do you think of this task? https://lab.apertus.org/T1018

Would that be interested for any of you to do?

Bertl_oO: dealt with ZNC afaik

supragya_: let's take this private, it should work

Bertl_oO... Wait let me reconnect

BAndiT1983 what green are you talking about?

green, like green sensel ;)

I was thinking the other way

so, red starts a 1,1 makes sense.

green1 starts at 1,1 -> cross ?

yay!

ok

bilinear debayer walks over unknown parts, that's why the offsets are in that way

and finally blue is 2,2 -> diagonal cross, understood!

stop, no

green is 1,1

*you start counting

on the code G1 is 1 1 and G2 is 2 2

possibly it was done for processing more than one value

Hmmm

the BilinearFilterRed starts with a diagonal cross for the starting position, then goes 1 unit to the right and does top bottom and finally goes 1 unit down from the starting point and does the left right .

I am starting to see why was I having difficulties with the algorithm

feasible*

some data has to be calculated through weighting, but we discussed it already earlier

now the method BilinearFilterGreen

Thanks for the explanation BAndiT1983 !

we can make this loops simpler, but hopefully without much performance loss

we could also have a variable for _size - _width

*that

_patternOffsets[0] = _width + 1;

_patternOffsets[2] = (2 * _width) + 2;

break;

you have to give me more hints, what is this snippet about?

that specify the starting points for first missing pixel of each color

Yeah

test 123

yeah, but that was me testing the account :)

Remember the pattern issue I had with the samples?

When I open the sample, it is noticeable RGGB, but the OCimage says it's GBRG. And when I force running with RGGB offsets, the output is correct.

It's still happening, I have to do the unit test to conclude for once and all if it's a issue with my algorithm or with the TIFF loader or with the sample

Did it happen to you in the past?

okk! I will go dinner now!

have a nice meal

Hi all..

just joined, because I'm very interested with this project.

I have some skill in hardware and software development.

you can start at lab.apertus.org, if you haven't visited it yet, to get some overview of current hardware tasks

hey tjstyle!

Hi Bertl_oO !

no problem, we are mostly not-native speakers here

so the hardware should be there, just missing software ... which doesn't mean that you might not create your own design for that

yes, a first version was designed and populated a few days ago

that info was available in the wiki?

ok then, please give some detail about existing IMU CSO board.

https://wiki.apertus.org/index.php/Beta_CSO

http://files.apertus.org/HARDWARE/AXIOM/BETA/CSO/

the IMU one is obviously the one we are talking about :)

so I can't do anything in software side :(

which is the only one for the CSO

i.e. it can be accessed via I2C or SPI from the zynq PS side

was play with much IMU chips in drone world, so have some knowledge for what a good chip for certain need.

TofuLynx, any new questions?

IMU CSO only have SPI & I2C interface?

almost all good GPS chips that I know is using uart protocol

None yet. it's going well :)

samples*

so for that I would suggest to use the SHIELD interface instead

http://files.apertus.org/HARDWARE/AXIOM/BETA/SHIELDS/

so maybe work on a GPS shield then instead of a CSO IMU :)

we haven't tested it yet, it might be true but we'll see

there shouldn't be too much difference between CSO and SHIELD are regarding pressure

but as I said, we haven't tested it yet

did you manage to get a voucher?

not yet have info about that

Bertl_oO, what about heat pipe, so the fan can be put elsewhere?

I was late for that, I just found this project after crowd funding was over.

tjstyle: yeah, that's unfortunate ...

like I will find microzed locally, then buy a power board, main board, etc.

and you can always build those boards yourself if you have some soldering skills

yes, that with the sensor is definitely the most expensive part

also now I assembling my handmade product.

nice

thank you for that.

no probem, and if you have any technical questions, just ask me here on IRC

I noticed a "issue"

in the board photo above?

tjstyle: yep

is it normal?

looks like an audio board

Bertl_oO : My question probably applies to you too, is it normal that green1 sensels are darker than green2?

Do you know a dng sample of axiom that is okay?

no, but if you ask se6astian nicely he might be able to take a snap with all the stuff turned off

Oh Okay, thanks

monochrome and color sensors are basically identical and it would be weird if the sensel had different sensitivity based on row/col locations

it would also be weird to have different filters for G1 and G2

well the green1 sensel would by some reason be different to green2 for more range or something, no?

but i am no hardware expert :)

so you probably can take your own snaps and play with all the various parts

I don't think that is that. I have inspected thorougly and debugged

I will debug it more

maybe there are some tags with related settings, can't say for sure, cause this thing is full of surprises

it should be easy to split the raw12/raw16 into four images and check them against eachother

CFA Repeat Pattern Dim : 2 2

what is this?

yes, kind of, the MachXO2 (Lattice) is used as a routing fabric, so you have several options there depending on the configuration

tjstyle: it is a full featured FPGA (the MachXO2) so you can (pre)process the data from/to the CSO there

it also has some conenctions to the Zynq (Xilinx FPGA) which can be used for communication

and pattern 2 is GBRG

tjstyle: so one way would be to pre process the IMU data in the MachXO2 and then send metadata packets to the ZYNQ

I cannot find any material regarding that

ah ok, got it.

so the sample as the wrong pattern

has*

https://www.adobe.com/content/dam/acom/en/products/photoshop/pdfs/dng_spec.pdf -> page 23, very interesting

bayergreensplit

tjstyle: there is a GSoC task which deals with an advanced communication protocol between MachXO2 and Zynq to simplify this

https://www.awaresystems.be/imaging/tiff/tifftags/privateifd/exif/cfapattern.html

TofuLynx, can it be, that the pattern is set fixed in the loader?

CFA Pattern : [Green,Blue][Red,Green]

you should also bookmark the page, as the info there is very important when dealing with TIFF/DNG

Bookmarked!

Hmm if the sample has the wrong pattern in the tag, why did that happen?

the same as exif, GBRG

yep

https://www.awaresystems.be/imaging/tiff/tifftags/cfaplanecolor.html

darktable?

ok!

darktable is just an opposite of the light table, which is used by photographs and pre-print companies

without any profile, it's has a magenta hue to it

that's normal, at least for now, as the conversion has some problems, maybe we can fix it later

i.e. the dng conversion didn't use the correct RG/GB pattern because even and odd rows were swapped

so, do you think that the sample does have in fact a wrong pattern?

Bertl_oO , can I send some prints?

(and paste me the url here or in private)

here is it

that looks like swapped lines :)

(which is because they are actually G1 and G2 :)

how do you suggest to compare?

Blue and Green Channels are similar in the darktable

I have to go now. Tomorrow I will continue this :)

xD