#apertus IRC Channel Logs

Good morning

Hey folks!

very good, have you succeeded to extract the output back to your host machine?

what is the error message?

I wouldn't exptect the QEMU tftp server to support put requests

looking at the code, this is easily confirmed:

if tftp is working, so must the network also, or not?

i.e. the connection between the virtual network card and the qemu network stack has to work for this

Yes I can do get requests

Seems like TofuLynx too had problems installing beta software

(rip my grammar)

Yeah, install the old version, with the old Kernel, the newer Kernel still has issues

Yeah

Without problems at all

BandiT, my C++ challenge was completed a week ago... or maybe beofore that... I was trying to get beta softwar to work... but it was ready..

worked on it

i know, Bertl sent me the link, have to add it to the c++ challenge task, so we can review it

Here is it supragya :)

you can push it git, review will be done as soon as someone has time, have to finish some personal stuff before that

BAndiT1983: how do you know I run it on the beta firmware?

Tofulynx, can you provide me with commit number

nvm, got it

blank? there are 3-4 applications, partially with features (see screenshots in wiki or lab)

What are you planning to work on Tofulynx?

The same as you plan, I think xD

Were you successful with the old Kernel?

Sorry was away

Bertl, what is this discussion about tftp server? axion-beta-software related?

it is about the axiom beta firmware, to test the code on the firmware you can use a tftp server to get files from the Host

This beta firmware, the new one or old one ?

were you successful in running it on old firmware?

that cpp one?

I am asking, were you successful in running the debayering program on old firmware?

k

I am mainly interested on T722 OpenCine: Raw Image Debayering Methods

And you?

I would like to work on OpenCine more than any hardware related things but really need to look

Wishes of the best luck for you!

most stuff for the beta i'm testing under manjaro, as it is archlinux derivative

Bandit, after building OC

there should be 2 more, at least

well, no

used Qt gui and cli both

not using qtcreator

will try it later on a fresh linuxmint vm, to ensure it still works, but was adjusting OC some time ago, had no problems

[ 6%] Built target MemoryPool

[ 8%] Built target OCcore_autogen

[ 56%] Built target OCui_autogen

[ 84%] Built target ProcessingTest_autogen

[100%] Built target ProcessingTest

supragya: no, the tftp was just QEMU related

much is commented out which looks like is the culprit

my bad, sorry, was improving processingtest for gsoc a bit and then forgot about this

can you do this so i can try tinkering with OC code?

nothing bigger, this 2 lines are the problem

and if you run cmake again, then it should find it, then make and you should be good to go

which ones? i don't have time at the moment to take a deeper look, have to fix stuff elsewhere

it was a cache problem (feel stupid now)

it can transfer data from a drive to defined place, thumbnails must be broken, because i've deactivated FFMPEG for now

FFMPEG was deactivated by me, because there were build problems on windows

FFMPEG is a library

for MPEG and video related things...

used it to get preview for mov files

software seems in it's too new a phase... am i wrong?

you have to insert virtually or real an sd-card or similar

or where did the screenshots in wiki and lab came from??

it works with removable drives, at least it was the filter for it

but i've used linuxmint then

OCbackup seen the changes and shown it in the list at top-left corner

which editor? ocbackup is a module, the biggest one, compared by implemented stuff

does not look like OCBackup

this aws what i was asking :)

not implemented as of now right?

a playground for visual stuff, i you want to call that so

vapoursynth is a successor to avisynth, this are frame server, where you use scripts to process video clips, decode, encode etc.

So like a video editor?

it's more a small IDE with CLI, you use mainly scripts

Can you elaborate an example of what I would use VapourSynth for?

for avisynth,predecessor -> http://avisynth.nl/index.php/Script_examples

just like you used lodepng's encode to encode png

think of a vieo editor software using vapoursynt

It seems awesome

it acts as a library/plugin for that software... the trick is to add to the vapoursynth thingy, OC's capability

BAndiT1983: correct me if wrong

no, it isn't working asa plugin

the application, which is not working with RAW data usually, opens that file, and vapoursynth reacts to the requests for frames and so on

so it would be: OCcore -> VP plugin -> VP frame server -> video editor

can be, i don't know adobe products that well

no

vapoursynth provides a file, which looks like simple video file

when the application requests data from it, frames, audio data etc., VP provides it, from different sources, in our case it should be OCcore lib, which will be accessed thorugh VP pplugin

yes

TofuLynx: http://avisynth.nl/index.php/Frameserver

?

But doesn't it take too much computational power?

we can also send just partial data, you remember our talk about half size or even quarter, similar to after effects preview?

i guess it should take more computational power than it being simple file in first place

in theory, the data is already loaded through OCcore, just the transfer is expensive, in my opinion

we can accelerate stuff through opencl, cuda, shaders or openmp, so the ball is on our site how we do it

Ok :)

this is different story, it's about how to gather different compositions in AE together, as they can be heavy on memory and CPU

so we don't have to write our own one, we would use VP

Makes sense

yeah

https://github.com/vapoursynth/vapoursynth

but we don't have to modify it, just to write a plugin, is not very difficult, tried it before, as preparation for that task

Ok!

supragya: thanks!

https://github.com/dubhater/vapoursynth-damb

but the structure is a bit questionable there

BAndiT1983: this probably isn't your area, but axiom beta camera can do the processing in the fpga right? Like a phone

Bertl is the expert

Hmm ok! :)

Bye! and thanks BAndit1983 TofuLynx, gtg

off for now ... bbl

Hello. I am Shubham Rath. I am interested to contribute in this organisation. I have knowledge of computer vision and I work with opencv and C++. I went through the ideas list. Can anyone guide me?

and welcome

Welcome!

what are you interested in?

we have several areas where you can contribute, e.g. OpenCine, a toolsuite for RAW data backup and processing

Do you use github or some other version control?

https://github.com/apertus-open-source-cinema

*which are

i think you mean gsoc, as there is a separate section for opencine and it has more than 2 ;)

okay :D

the idea page just contains some "suggested tasks"

or an idea that is mentioned on the lab

okay. Thank you. And I need to complete the qualification task first, right?

and you have to do the C++ challenge before applying

okay. Thank you :)

yes?

have implemented it partially, there is still a lot to optimize, but don't focus on that first

Doanld E. Knuth

Does Axiom Beta

wait

like amateur users

ok!

that seems great! and what about the output of it?

which output?

it does both, HDMI is there and so on, but se6astian has more details

and simple .h265 videos?

currently we write still images in RAW12 format in the AXIOM Beta: https://wiki.apertus.org/index.php/RAW12

ok!

not sure how farfetchd is that idea, maybe doesnt really have a point

you certainly can do that, but what would the benefit of an NPU be for the typical cinema movie maker?

note that the Beta is flexible enough to be adapted for almost any purpose as long as the hardware permits it

Yeah, If I had the knowledge of FPGAs and NPUs and hardware, I would try something like that, a AI Powered camera, but there's really no point in that and i am from informatics engineering xD

wow xD

The modularity

@sr6033 No! I am a student interested in contributing and participating in GSoC. OpenCine is my target!

namely the task T722

Hi! sr6033, welcome

me?

not upto any real convo

Hello @supragya :)

me -> https://www.apertus.org/user/7

in general the best approach is to just ask a question here

(often with an idication on the quality of the answer or a hint who might know more)

that is something only BAndiT1983 is able to answer

BAndiT1983|away: is working on OC

So does OC decode anything RN? The OCcore is pretty empty for https://lab.apertus.org/T763... Cannot find any RAW

RN?

Feel free to use, let me know if you need a license to go with it

very nice :)

BAndiT1983|away: will be able to confirm

That sounds important

felix_: did you test the reworked SDI module?

would be nice to know if it now works as expected (when you find the time, not urgent)

i'll see if i have time for that tomorrow

IMHO the only reason for intermittent working (as described) could be the razor beam connectors

so I simply re-did all of them

felix_: good to know!

these photos look amazing, how were they shoot?

2 led desk lamps with a sheet of paper as difusor each, 1 mobile phone with another piece of paper as diffusor, 2 pieces of paper as background and a good camera with tripod. and then a bit post-processing with gimp. iirc DrLuke also made a photo of that setup

Thanks felix!

Thanks, the setup was a bit ghetto, but I tried my best with the stuff I had available :P

Maybe I will give a shoot at OpenCL cores?

That setup is really improvised! Awesome!

Ok! Will check it :D

You give preference to OpenCL, right?

DrLuke, did you use an axiom beta kit to capture the photo or was other machine?

with kit lens

what is kit lens?

lens supplied with the camera

TofuLynx: the lens that comes with the camera when you buy it

they are ok quality

but the high-end lens produces much sharper images, you could zoom in two or three times more before it gets blurry

I always thought it's just placebo, but I've once tried a 3000€ lens on my camera and holy crap it's like a whole different camera

3000€ tho

TofuLynx: https://www.amazon.de/Canon-EF-70-200-USM-Objektiv/dp/B0033567D8/ref=sr_1_8?ie=UTF8&qid=1519508161&sr=8-8&keywords=canon+L

yeah

That's the kind of glass you'd buy if you made your living with photography

Understood!

Heh yeah... if only :P

but also how to load just half the size or quarter and how the performance affected

what is your definition of fast?

also, do tell me one thing

BAndiT1983, isnt MagicLantern a firmware while OC an app?

linearization is usually based on the additional data from sensor, if i remember correctly

ML is an extension for canon firmware, they have own data format, have implemented a decoder for it in OC

for example one could use look up table for performance increase

yes, but aperture come from lens, iirc

So how would the thought of workflow be? Obv, ML will be frameClient through VS using OC plugin, which will decode .... ? And where will this run? In camera..? Sorry didnt actually got the full picture

ML has it's own video format

OC is for the PC, it's a suite to process data from the camera

Bandit, you told me that the objective of the occore plugin was to bring to VS something which will have OC as backend to process the RAW etc. Something like ML. I am just trying to figure what you meant.

Quote-"for software which does not support raw videos or for example MagicLantern, we wanted to provide a plugin for vapoursynth, so we can access OC as data handler"

good night

BAndiT1983: TofuLynx : I think opengl would be a better choice for debayering, because problems that map nicely to the graphics domain are generally better solved with opengl than opencl

anuejn, are you sure?

is the bayerfilter a pixel-perfect thing?

se6astian: You commited some file with a space in the path :P. this breaks my setup :rofl:

moreover, wouldnt it be smart do delete all the php stuff from the github repo and replace it with the stuff you put inti the dir with the evil path?

DrLuke: yes it has to be done per pixel

DrLuke: https://anuejn.github.io/batic/

anuejn

but excelent for displaying the result

TofuLynx: Why do you think it's unsuitable?

"OpenGL may place your video within the editing interface and make it play, but when you throw color correction onto it, CUDA or OpenCL will do the calculations to alter each pixel of the video properly."

opengl is optimized for pushing around pixels

this is pretty classical opengl territory if you ask me

and "pushing around pixels" includes the debayering process and linearization and stuff?

makes sense

TofuLynx: This is no different to any of the post processing shaders that are present in any modern game engine

yeah, I understand

to display it to the user as fast as possible

That's OpenCine's goal

Ok! Thanks!

I think it makes sense, will have to get a complete explanation ready

TofuLynx: None

Hmm

anuejn: btw: glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR_MIPMAP_LINEAR); will make the preview look much better when zooming out

VJ?

DrLuke: the zooming is just the browser transforming a canvas object, which i cannot influence

was a quick hack at the cost of aliasing

that's great!

Jup, it's fun :)

DrLuke: so you are into demoscene stuff?

I planned to release my first demo this year at revision, but time got ahead of me

But yeah, I admire everything they do

TofuLynx: one of the biggest demoparties

basically you go there to finish and release your demos

BAndiT1983: available? Ping!!

maybe :)

supragyaraj: IRC bridges space _and_ time ... just go ahead and ask what you want to know, you'll get your answer sooner or later

anuejn: Oh neat, I'll be VJing there aswell

Don't tell anyone, but neither am I! :P

Anyway, time for sleep, good night!

see you

Great to meet you!