added some cleanup code
remove the largest files at the end
updated QoE.pl to new chunk modes
upgraded psnr.sh to support new chunk modes
started to adapt psnr.sh to the new "chunking mode" parameters
added fixedsize and minsize methods to chunker
added k frames / chunk chunkizer
print remaining frames as last chunks
added chunk size parameter to Chunker
moved chunk printing to separate function
fixed an annying error where the resulting number of duplicates was -1
removed duplicate evalpsnr line
changed cutter to cut based on frameloss also when mode 0 is used
print PSNRencoded with encode PSNR and SSIM
fixed psnr calculation for the encoded video (without chunk losses)
calculation is done right after encode, before chunkization. This couseda problem, since peer_qoe was looking for chunks.txt.TODO: better have a separate method like src_qoe for this.
fixed refiller crash if number of dups was too high.
Max number of duplicates was set too low, and there was no error message.
added new "chunk" method to QoE.pl
added new "chunk" method and chunker support to psnr.sh
added chunksize/method option to chunker
with -2 it creates ipb chunkswith 0 it creates GOP sized chunksothers not yet implemented
initial support for chunks.txt and ipb chunkization
changed chunker output to the chunks.txt format defined in the simulator
added gop support to chunker
renamed files in Chunker
added Chunker that creates chunks.txt from a video
added new cutter working mode to remove frames
adds a new working mode to the psnr-toolscutter (chunk size -2), reading the list of lost frames generated by"lf" and removing them from the stream (this can also be used toimplement the "1frame=1chunk" strategy advocated by polito).
LostFrames gets as input the output of analyser and a list of lost chunks, andgenerates as output a list of lost frames
fixed refiller error due to decode order
encode could return 0 in some cases. If this happened at the beginning,refiller was segfaulting.
fixed a refiller issue, which I think is triggered by H.264+Bframes (Luca)\
The problem is that when a chunk is lost, the first correctly receivedframe after the chunk cannot be decoded correctly (because of missingreference frames). libavcodec does not apply any error concealing here,...
fixed a logparser bug, which ends up in duplicating the wrong frames (Luca)
The total number of frames in unchanged.
changed make to $(MAKE) in Makefile
to resolve this warning:make1: warning: jobserver unavailable: using -j1. Add `+' to parent make rule.
added some usage info
added play support with ffplay
FFPLAY must be defined to use it
fixed makefile adding .PHONY
[PATCH] Remove the "gop" hack... Now 1chunk=1gop is chunksize=0
create temporary directory if it doesn't exits
minimal README added
fixed overwrite problems with ffmpeg
previously ffmpeg was defined with -y (answer yes to overwrite questions).Now -y is part of the script itself.
Merge commit 'abeni/smart_chunkisation_merged'
fixed bash =~ error
Now it works both for bash 3.1 and for 3.2.48seehttp://books.google.com/books?id=Qu6H8fnAvGoC&pg=PA123&lpg=PA123&dq=bash+rematch+3.2&source=bl&ots=CgiP4erH20&sig=4vbkbdx6o-nN1ZQWnzfjOXo3x8s&hl=en&ei=wfuTSsv8N4je-Qa62NCxBg&sa=X&oi=book_result&ct=result&resnum=1#v=onepage&q=bash%20rematch%203.2&f=false...
fixed error in desc file copying by removing the desc file.
alternatively, -f could be used
do not suppress error messages
previously there were some 2>/dev/null redirections,since ffmpeg writes a bunch of logs there. Since also useful errormessages are there, it was re-enabled.
Build system fixes...
Support new cutter, and do not hardcode ffmpeg
Fix support for interleaved frames
Fix support for fixed and variable chunk size
fixed ssim rounding problem
evalvid was printing results up to a precision of two decimal digits .... OK forPSNR, but not for SSIM which has a theoretical scale of [-1...1] and areal scale of sg. like [0.8..1]Changed to 6 digits
Merge branch 'master' into smart_chunkisation_merged
Conflicts: Cutter/input-avs.c Cutter/main.c
Smarter chunkisation: 2 chunks = 2 gops, interleaved frames
fixed bash regexp error.
Bash regexp behavior changes from version to version. Seehttp://books.google.com/books?id=Qu6H8fnAvGoC&pg=PA123#v=onepage&q=&f=falseHope this works now with all 3+ versions.
Make the chunk type selectable at runtime
autodetect video parameters, or in case of yuv require .desc file
changed logparser to match based on frame size
pts based matching was not working with x264
added x264 support link support in Makefiles through the X264DIR variable
only needed if ffmpeg was compiled with x264 support
corrected mismatch of +-1 frame in refiller
Refiller was duplicating the frame before 2 before, not 1 before.
handle XSIZE, YSIZE and FPS in psnr.sh
added ssim calculation
added GPL notice
added coder parameters as external parameter to scripts
fixed chunk buffer: it was too small
added psnr without chunk losses to output
changed TMPPREFIX to be an external env variable
changed script paramters to handle init separately
added scripting friendly PSNR output
fixed node column ID in chunk_lost
fixed PSNR_TOOLS env option handling
fixed segfault with max chunks lost too low
quick fix for psnr-tools folder
fixed error while taking psnr average from log
added "every" parameter to QoE script
just a bit of output formatting
fixed error passing chunksize parameter
calculate base que, as well as GOP based and fixed chunk size qoe
Added some output to init
It is commented since at the end it was not needed. Instead ofthis we run calc for the source.
added scripts to process simulator output and compute quality indicators
changed script output to be on stdout
separated init from calc
minor fix in script params
script calculates bytes, chunks, chunkloss and psnr
added statistics output to cutter
added first version of PSNR evaluator script
added some readme files
added a makefile for logparser
fixed error with last newline in analyser's output
-2 error was reported because of the newlize at the end of the input filethis is just a quick and dirty workaround!
fixed init values for some variables
error (-1) was reported due to non initialization of these variables
changed PSNR value in case of perfect match (MSE=0) to 255
more realistic than the original value of 0
added program to reconstruct the video's missing frames by duplicating previous ones
added program to remove chunks from a file
added program to compare outputs of analyser and find lost frames
added program to analyse the video sequence