#apertus IRC Channel Logs

BAndiT1983|away, able to write headers, made AVIEncode class here: https://github.com/supragya/libfuse-FrameServer/tree/master/AVIWriter

BAndiT1983|away, I have a few questions regarding some of the headers and format of frame

Students application deadline extended to 27?

nmdis1999, It had always been till 27

supragya, will take a look today at evening at earliest, when i'm on the train back home

Does anybody have some papers regarding SHOODAK debayering algorithm?

Thanks!

give me a day

BAndiT1983|away, malloc's more of a fallback to C. Had problems when I had to selectively give functions in stat structure for FUSE to work. G++ complains if assigned functions out of order. But i guess new and delete can be used in AVIencode.

TofuLynx: I found documentation in German about Shoodak

I will transfer (translate) everything to a wiki page

german DOP: Stefan Schuh (http://www.imdb.com/name/nm0775939/) invented it and is collaborating with us

What should I do next, Test the same on zynq platform, or move on with other tasks?

Hi Arun___ , which clg?

GgsIpu University campus

nice

(which makes the code hard to read)

Arun___: e.g. in Receiver.vhd why are the generics and the ports on different indentation level? why is std_logic in line 14 on a different position than in line 15?

in general, put spaces around operators like :=, +, etc

Ok ,On it

VHDL provides elsif and else for this purpose

processes without a label/name are also a bad idea

the registers could also become an array

Indentation maybe caused because i edited the whole in xilinx ise

well, it's fine if you use Xilinx ISE or any other IDE, but when you're done, you want to clean up the code :)

where did you get the lfsr.vhd from?

Arun____, happened to all of us (that embarassment) XD!

@supragya: I am not from GGSIPU, I am from IGDTU it's not under IP uni anymore xD

supragya: https://wiki.apertus.org/index.php/Shoodak

se6astian, clean and self contained wiki! Thanks.

stephan has an untested idea for a Version 2 of Shoodak as well already

but I guess focusing on Version 1 first and once that is covered trying more experimental stuff is a good approach

completely different approach

you cant get enough can you? :)

stephan worked with a german software developer who wrote a small tool to process still images with shoodak

:) - I have Object Oriented Software development exam tomorrow! What do you think I would focus on... that? XD

se6astian, I guess yes!

same syllabus

but you have to do something with it then :)

but for now debayering is Tofu's dept

se6astian, SHOODAK v2 looks more like Nearest sensel based debayering to me

or can I write one?

there is no code for V2

se6astian, sure! will do

if we find out his V2 method already exists and has a different name than thats also a conclusion :)

i am referring to V1

I think the description misses to strenghten the randomness factor

gotta go

danieel, for randomness in V1?

random influence - i pick up this as random per each pixel?

plus there are examples of random picks of triplets below, but no text to tell why or how to use it

danieel, I guess it is clear as to what triplets do

That triplet decides what sensel has more influence

zipper artifacts are regular

but you use them with at most 1 of 8 chance, so your zipper gets just 1 teeth of 8 (ideally)

This will make the line jagged of course

do you have any easy c code for it?

se6astian says he has one made by some german Software Developer. But I guess the task is not that hard

can you point me to your existing code? danieel

well I would like a simulation system to work with. Would really like a look at demosaic.h and matrix.h.

your task is to go from the raw file, I can not help you with that

give me some time. Perhaps a day

have luck!

which exam tmrw

Hello, BAndiT1983

danieel: supragya : implemented the random and anti zipper parts of shodak :)

can be used with https://anuejn.github.io/batic/

BAndiT1983, I think AVIEncode will be done soon. However

and is 'idx1' necessary?

anuejn, thank you! will have a look (net too slow to open your site)

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

have to check some things, as i don't have full overview of what was done in last days

i have no matlab, so cannot check it

.m files?

https://github.com/apertus-open-source-cinema/debayering/tree/master/Debay-O-Mat

*it

which namespaces do you mena?

code in general. namespace AVIEncode to have it all in one place

structs in class a no no?

my preference is clear structure, single responsibility per class etc.

not this one

i know that you have ported it from there, but this approach kills performance

I have not

I dont even read file wright now

I am just writing headers

are you sure you did a git pull on [libfuse-FrameServer] and looked at AVIWriter/ ?

If you are busy don't worry

i'm not busy, just riding train back home and trying to get an overview

looking forward to it

static struct fuse_operations fuse_example_operations = {

.open = open_callback,

.readdir = readdir_callback,

sorry, unimplemented: non-trivial designated initializers not supported};

tried this: https://stackoverflow.com/questions/31215971/non-trivial-designated-initializers-not-supported, no help

BAndiT1983, I guess we have to assign a function to every stat function as it is given:

Assign to .name=value fields

Include all fields in the assigment

*?

AVIWriter

I have setup a readme

It is not

only headers are written

but it was not playable, so wrote from scratch

that is in specifications

i have seen it, definitions are from microsoft, if i remember correctly

Writing 'db' format (uncompressed) wont be hard, but we just have to get all headers correct

can you explain dwInitialFrames?

https://msdn.microsoft.com/en-us/library/ms899423.aspx

for us it should be 0 for initial frames

we need very basic AVI stuff without much junk, plain and simple header, then a list of frames

will do some adjustments to your code and try it also myself to get more knowledge

hope that i will get a bit of time in the next days, as i have a lot to do with moving my stuff to new apartment and especially registrations in the new city, so i can park my car, needs special permission in the new area

we can incorporate the downscaler from TofuLynx at some point, when we get to playing back DNG sequences

the downscaler is ready?

ofc... but we need it eventually

also would extract some general methods to a general helper class, like 12->16bit or 14->16bit conversion etc.

in downscaler

we should also add some tests, i suggest catch2 framework, as it's a single file

unit tests, to verify changes in code, so we know when something breaks

what is the scale?

sure

if we have working file, we would take it as golden master, after changes new file will be written to disk, then golden master loaded and both binaries compared

Hey

Just read the logs

hi TofuLynx

Will check this night or tomorrow morning :)

gspot or mediainfo under ubuntu

like a RIFF tag editor

and it is hard

but with some inspector we would know precisely what is missing

match with actual AVI

seems like that, as the video player just complained about streams

Hey! Sorry for replying later, I was working on the tool :( Tommrrow's electrical science. After tomorow I'll be able to focus on my work here.

we can have a synthetic image

then we can check for the actual frames

BAndiT1983, thanks for mediainfo, will try to see what it can do

ah, forgot that you are on another distro

and it left me hanging

so temporarily installed 18.04 in vmware first

doesnot seem like what we need

I tested on my system, same OP

https://mediaarea.net

maybe the headers are not filled right

i dont understand

BAndiT1983, you can check hex editor

hdr1 list is okay

not written anything further

where is the AVI stuff? (why avi today?)

whats that

for OC

it's a prototype we are building

using readcallbacks

so you have to properly handle the seeks, right?

yes, handle the seeks

we can accelerate by reading image data only at fixed offsets

but there are also many other ways to do acceleration

both need multithreading and maybe HW acceleration later on

what do you mean with success?

seen the header, will analyze a bit later in depth, but first i will adjust the code

nope, just locally first, can still fork later

then get a faster drive

the real bottleneck will be memory bw limit

depends on purpose

but if you speak of like using or not using o_direct and zerocopy.. yes, do that first

the point is, to get most out of given hardware

not to hog it up with unnecessary code bgs

know a lot of people who cannot afford or do not want to buy new machines

I guess it is because the softwares are more buggy

it'S just a part of the problem

http://www.file-recovery.com/avi-signature-format.htm

profiling still has to be done, but first some working code has to exist, before even considering fiddling with bits

slowness is caused by nesting technologies, virtualization and too high level languages

have to look at size obv

it can be done by memcpy for special cases

ah, just scrolled down and seen that, but have to test if it's not better to do it directly in header code

as it is an avi writer, we shouldn't submit data from outside, just set fixed data right there in code and supply modifiable data like frames

I have tried to keep the clutter out of the way

however we still need the width, height and framecount info

will evaluate it, but there are too many parameters

well, much of the headers are such that they need to be given explicitly

?

more like setFrameCount

?

e.g. setResolution(width, height)

supragya, is this header of yours really one which is used in general?

have a look at: https://msdn.microsoft.com/en-us/library/windows/desktop/dd318189(v=vs.85).aspx

and avih main header

and then movi

Stream Data ('movi' List)

so there is no strl and strh

just AVIMAINHEADER is done, not even AVISTREAMHEADER

but RIFF is structured by ID and size, afterwards comes the payload

RIFF filesize filetype(the ID) payload

RIFF 0000 AVI payload

I told you, i have not computed the size

ok, but that one is also important, even if fixed at first, otherwise applications don't have the right info

TIFF works also like that, ID and size, then payload of the tag

TIFF is near RIFF, T -> R XD

what i think is that we need a computing mechanism

that can exactly tell what info is needed when a particular part of file is referenced

the headers are to be precomputed and stored in memory

frames generated on the fly

http://www.fileformat.info/format/riff/egff.htm

you have a chunk, which consists mainly of ID and size, afterwards comes the payload, but it differs from yours

Ah I think you did not get one thing right

LIST and CHUNKS

CHUNK is ckID ckSize ckData

hence it is RIFF 0000 AVI payload

chunks are only featured when frames are written

see microsoft link

payload may have framedata and header info

not a big deal, but raw speed is our goal

Thus, I asked you to have a look before I move ahead

well I am a student

i will ask you in a couple of years again :D

had some arguments with old developers, even when their code was not doing the right thing, they refused changes

Today I had machine learning as exam

And we had to calculate 15 point distance matrix in 4th dimensional space

well euclidean distance were to be computed as sqrt( (x1-z1)^2 + (x2-z2)^2 + (x3-z3)^2 +(x4-z4)^2 )

for all x1 and z1 in that 15 points

like 100 of these in barely in an hour

and then to cluser them

printed circuit board

in exam

I wrote random numbers and made clusters XD

I don't prefer this system tbh

how so?

but it's also mainly about solving problems on fixed path and not about approach it in clever way

good night

gn

se6astian|away, BAndiT1983|away, Bertl_oO , I have redone T872, kindly give feedback on it :)

line 21, if(...){ -> if (...) {

yup

line 35, width ; -> width;

line 42/43 }\nelse{ -> } else {

shall I go on or do you get the pattern?

supragya, you should get an autoformatter

yes

there should be one already, but possibly you have to adjust some things

for C there is also 'indent' which does a pretty decent job

well done

I have done autoformatting

:D

i do it on regular base for java for whole directories

how is analyze feature of clion?

BAndiT1983, have a look at the structure of this attempt

we are talking about c++ challenge right?

tell me if this one is alright... used buffer this time, one file access

have to take a thorough look, but it's late here, so just a quick glance

Now I don't have to think of Bertl asking me to indent better, I have Ctrl+Alt+L XD

you should adjust formatter, so the braces start at new line

That's thing of preference right?

brace start

getting rid of sizeof(uint8_t) would allow me to have new char[width*height] which is better

supragya, you can also merge G1 and G2, no reason to separate them

I would like to have channel output for every sensel

green is green ;)

BAndiT1983, that true

what Bertl_oO said

you can also try to run cppcheck over it

but analyzer in clion should also work

it's a static analyzer which tells you about errors, problems and bottlenecks

s/out that/that out/

s/out that/that out/ ?

for those who have never used sed or vi, s/a/b/ replaces a with b

got it... not used vi for long

avoiding it like hell :P

BAndiT1983: but there is a GUI for it ;-)

using nano for terminal stuff

or without GUI?

sometimes

supragya, he is writing FPGA code without UI, so he is not far from becoming Neo

how does he do simulation?

;)

XD

I have hardware!

half-offtopic -> https://hackaday.com/2018/03/05/fpga-makes-ascii-video/

vigor? google gives different results

:D

+---------------------+

+------------------+--+

|

\\ |

X:\ (_/ \_) <------------------------+

\::\ \== /

\\/\::\ / / ( l

\/:::| `.' `

\/`-'`. > )

| /( .'

__||/_ \ `-'

whoa, next time you'll get kicked for spamming the channel

LS | # |# | - Front View -

gtg, good night

supragy -> http://www.asciimation.co.nz/#

thanks for the link

https://lifehacker.com/373571/watch-star-wars-in-text-via-telnet

so, enough off-topic, was checking oyur code

[ was checking oyur code] -?

check if the double loops, are really necessary

i assume yes, as i'm still trying to get the structure right in my head

let's continue another day

yup

glhf

gn

nn

off to bed now ... have a good one everyone!