A Dynamic Adaptive Streaming over HTTP (DASH) toolchain for Scalable Video Coding (SVC). This open source toolkit provides scripts for demultiplexing and re-multiplexing video content which has been encoded according to the H.264/SVC extension. It can be used together with the DASH/SVC Dataset provided here (http://concert.itec.aau.at/SVCDataset/).
This toolchain provides python scripts for converting videos encoded according to the H.264/SVC extension into several segments and several layers (spatial and quality scalability), allowing Dynamic Adaptive Streaming over HTTP for Scalable Video Coding. This toolchain is based on the SVC Demux/Mux Tool of Michael Grafl (see http://www-itec.uni-klu.ac.at/dash/?page_id=1366 for more information). This work was partly funded by the Austrian Science Fund (FWF) under the CHIST-ERA project CONCERT (A Context-Adaptive Content Ecosystem Under Uncertainty), project number I1402 (see http://www.concert-project for more details).
This README file will cover downloading, building and executing the scripts and programs provided within this github repository.
- Testing JSVM
- Analysing an H.264/SVC compatible stream
- De-Multiplexing an H.264/SVC compatible stream into DASH/SVC segments
- Re-Multiplexing DASH/SVC segments into an H.264/SVC compatible segment
- Supports spatial and coarse-grain quality scalability
- Parsing SVC MPD file and extracting the URLs of all segments and layers
- Temporal Scalability is not supported yet.
- Only coarse-grain Quality Scalability is supported.
- JSVM might not compile on all platforms.
- Python 2.7
- Python easy_install
- Python module: bitstring (if you can not install it, go to the Bitstring PyPy Website and use bitstring.py )
- CVS (required for downloading JSVM)
- build-essentials and cmake
- JSVM Reference Encoder
- libdash library for parsing MPD files (included in this github project, required for MPD parsing)
If you plan to use our work, please cite our paper:
Christian Kreuzberger, Daniel Posch, Hermann Hellwagner, "A Scalable Video Coding Dataset and Toolchain for Dynamic Adaptive Streaming over HTTP", In Proceedings of ACM MMSys '15, March 18 - March 20, 2015, Portland, Oregon.
@INPROCEEDINGS{KreuzbergerDashSVCDatasetToolchain2015,
language = {EN},
author = {Kreuzberger, Christian and Posch, Daniel and Hellwagner, Hermann},
title = {A Scalable Video Coding Dataset and Toolchain for Dynamic Adaptive
Streaming over HTTP},
editor = {Wei, Tsang Ooi},
talktype = {registered},
location = {Portland, Oregon},
talkdate = {2015年03月18日},
booktitle = {Proceedings of the 6th ACM Multimedia Systems Conference},
publisher = {ACM},
address = {New York, NY, USA},
pages = {213--218},
month = {mar},
year = {2015},
isbn13 = {978-1-4503-3351-1},
series = {MMSys '15},
doi = {10.1145/2713168.2713193},
keywords = {DASH; Dataset; Toolchain; Scalable Video Coding},
pdf = {https://www-itec.uni-klu.ac.at/bib/files/dash_svc_dataset_v1.05.pdf},
url = {http://concert.itec.aau.at/SVCDataset/}
}
This section describes the scripts for downloading and testing. Create a directory, e.g., SVCDemo, switch to this directory and follow the steps below:
Install required packages and download this git repository:
# Requires: python (2.7), cvs, git
# libdash requires:
sudo apt-get install cvs git-core build-essential cmake libxml2-dev libcurl4-openssl-dev
# python setup-tools (easy install)
sudo apt-get install python-setuptools
# python module bitstring
sudo easy_install bitstring
# if you want to play yuv files, you need mplayer or any other player that can play yuv files
sudo apt-get install mplayer
# get this repository
git clone --recursive git://github.com/ChristianKreuzberger/DASH-SVC-Toolchain.git
cd DASH-SVC-Toolchain
Let's build libdash or just call the script build_scripts/buildLibDash.sh:
(Note: libdash is required for parsing MPD files; if you don't intend to do that, you don't need to build libdash)
cd libdash/libdash
mkdir build
cd build
cmake ../
make
if [ $? -ne 0 ] ; then
echo "Failed building libdash";
exit -3
fi
# go back to the main directory
cd ../../../
Download and build the JSVM reference software or just call the script build_scripts/buildJsvm.sh:
(Note: JSVM might not compile on all platforms. We have successfully compiled JSVM under Ubuntu 12.04.)
# get JSVM reference software
cvs -d :pserver:jvtuser:jvt.Amd.2@garcon.ient.rwth-aachen.de:/cvs/jvt login
cvs -d :pserver:jvtuser@garcon.ient.rwth-aachen.de:/cvs/jvt checkout jsvm
cd jsvm/JSVM/H264Extension/build/linux
make
if [ $? -ne 0 ]; then
echo "BUILDING JSVM FAILED!"
exit -1
fi
# see if the JSVM tools exist
cd ../../../../bin/
ls
# test bitstream extractor static
./BitStreamExtractorStatic
# go back to the main directory
cd ../../
Test JSVM by decoding a H.264/SVC video from our dataset:
# let's test JSVM by downloading a H.264/SVC video from our dataset (svcseqs subfolder)
JSVMPATH=$(pwd)/jsvm/bin
export PATH=$PATH:$JSVMPATH
# download a video
wget http://concert.itec.aau.at/SVCDataset/svcseqs/II/bluesky-II-360p.264
BitStreamExtractorStatic bluesky-II-360p.264 > bluesky_test.txt
diff bluesky_test.txt tests/bluesky_II_360p.txt
if [ $? -ne 0 ] ; then
echo "TESTING JSVM (TEST 1) FAILED!"
exit -2
fi
# try decoding it
H264AVCDecoderLibTestStatic bluesky-II-360p.264 bluesky-II-360p.yuv > bluesky_decode_test.txt
diff bluesky_decode_test.txt tests/decode_bluesky_II_360p.txt
if [ $? -ne 0 ] ; then
echo "TESTING JSVM (TEST 2) FAILED!"
exit -2
fi
# remove the files that we just created
rm bluesky_test.txt
rm bluesky_decode_test.txt
rm bluesky-II-360p.264
rm bluesky-II-360p.yuv
echo "TESTS DONE!!!"
Congratulations! You have successfully downloaded and built libdash and the JSVM reference encoder. In the next subsections you will try to decode a DASH/SVC segment with our toolchain.
Assuming you are in the DASH-SVC-Toolchain directory, follow these steps: (Note: This step requires JSVM binaries)
# make sure JSVM is in PATH
JSVMPATH=$(pwd)/jsvm/bin
export PATH=$PATH:$JSVMPATH
cd decode
# Download init segment
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/II/segs/720p/bluesky-II-720p.init.svc
# Download segment 0, Base Layer
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/II/segs/720p/bluesky-II-720p.seg0-L0.svc
# Download segment 0, EL 1
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/II/segs/720p/bluesky-II-720p.seg0-L1.svc
# Download segment 0, EL 2
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/II/segs/720p/bluesky-II-720p.seg0-L2.svc
# call svc_merge.py and create the yuv files for the segments
# svc_merge.py for base layer only:
python svc_merge.py bluesky-II-720p.seg0-BL.264 bluesky-II-720p.init.svc bluesky-II-720p.seg0-L0.svc
H264AVCDecoderLibTestStatic bluesky-II-720p.seg0-BL.264 bluesky-II-720p.seg0-BL.yuv
# svc_merge.py for base layer + EL 1:
python svc_merge.py bluesky-II-720p.seg0-EL1.264 bluesky-II-720p.init.svc bluesky-II-720p.seg0-L0.svc bluesky-II-720p.seg0-L1.svc
H264AVCDecoderLibTestStatic bluesky-II-720p.seg0-EL1.264 bluesky-II-720p.seg0-EL1.yuv
# svc_merge.py for base layer + EL 1 + EL 2:
python svc_merge.py bluesky-II-720p.seg0-EL2.264 bluesky-II-720p.init.svc bluesky-II-720p.seg0-L0.svc bluesky-II-720p.seg0-L1.svc bluesky-II-720p.seg0-L2.svc
H264AVCDecoderLibTestStatic bluesky-II-720p.seg0-EL2.264 bluesky-II-720p.seg0-EL2.yuv
# use mplayer to playback the three yuv files
mplayer -demuxer rawvideo -rawvideo w=1280:h=720:format=i420:fps=24 bluesky-II-720p.seg0-BL.yuv -loop 0
mplayer -demuxer rawvideo -rawvideo w=1280:h=720:format=i420:fps=24 bluesky-II-720p.seg0-EL1.yuv -loop 0
mplayer -demuxer rawvideo -rawvideo w=1280:h=720:format=i420:fps=24 bluesky-II-720p.seg0-EL2.yuv -loop 0
# remove the files we created
rm *.yuv
rm *.svc
rm *.264
cd ..
Assuming you are in the DASH-SVC-Toolchain directory, follow these steps: (Note: This step requires JSVM binaries)
# make sure JSVM is in PATH
JSVMPATH=$(pwd)/jsvm/bin
export PATH=$PATH:$JSVMPATH
cd decode
# Download init segment
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p/bluesky-III-1080p.init.svc
# Download segment 0, Base Layer
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p/bluesky-III-1080p.seg0-L0.svc
# Download segment 0, EL 1
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p/bluesky-III-1080p.seg0-L1.svc
# Download segment 0, EL 2
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p/bluesky-III-1080p.seg0-L2.svc
# Download segment 0, EL 3
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p/bluesky-III-1080p.seg0-L3.svc
# call svc_merge.py and create the yuv files for the segments
# svc_merge.py for base layer only:
python svc_merge.py bluesky-III-1080p.seg0-BL.264 bluesky-III-1080p.init.svc bluesky-III-1080p.seg0-L0.svc
H264AVCDecoderLibTestStatic bluesky-III-1080p.seg0-BL.264 bluesky-III-1080p.seg0-BL.yuv
# svc_merge.py for base layer + EL 1:
python svc_merge.py bluesky-III-1080p.seg0-EL1.264 bluesky-III-1080p.init.svc bluesky-III-1080p.seg0-L0.svc bluesky-III-1080p.seg0-L1.svc
H264AVCDecoderLibTestStatic bluesky-III-1080p.seg0-EL1.264 bluesky-III-1080p.seg0-EL1.yuv
# svc_merge.py for base layer + EL 1 + EL 2:
python svc_merge.py bluesky-III-1080p.seg0-EL2.264 bluesky-III-1080p.init.svc bluesky-III-1080p.seg0-L0.svc bluesky-III-1080p.seg0-L1.svc bluesky-III-1080p.seg0-L2.svc
H264AVCDecoderLibTestStatic bluesky-III-1080p.seg0-EL2.264 bluesky-III-1080p.seg0-EL2.yuv
# svc_merge.py for base layer + EL 1 + EL 2 + EL 3:
python svc_merge.py bluesky-III-1080p.seg0-EL3.264 bluesky-III-1080p.init.svc bluesky-III-1080p.seg0-L0.svc bluesky-III-1080p.seg0-L1.svc bluesky-III-1080p.seg0-L2.svc bluesky-III-1080p.seg0-L3.svc
H264AVCDecoderLibTestStatic bluesky-III-1080p.seg0-EL3.264 bluesky-III-1080p.seg0-EL3.yuv
# use mplayer to playback the three yuv files
mplayer -demuxer rawvideo -rawvideo w=640:h=360:format=i420:fps=24 bluesky-III-1080p.seg0-BL.yuv -loop 0
mplayer -demuxer rawvideo -rawvideo w=1280:h=720:format=i420:fps=24 bluesky-III-1080p.seg0-EL1.yuv -loop 0
mplayer -demuxer rawvideo -rawvideo w=1920:h=1080:format=i420:fps=24 bluesky-III-1080p.seg0-EL2.yuv -loop 0
mplayer -demuxer rawvideo -rawvideo w=1920:h=1080:format=i420:fps=24 bluesky-III-1080p.seg0-EL3.yuv -loop 0
# remove the files we created
rm *.yuv
rm *.svc
rm *.264
cd ..
Assuming you are in the DASH-SVC-Toolchain directory, follow these steps: (Note: This step requires JSVM binaries)
# make sure JSVM is in PATH
JSVMPATH=$(pwd)/jsvm/bin
export PATH=$PATH:$JSVMPATH
cd decode
# Download init segment
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p-temp/bluesky-III-1080p.init.svc
# Download segment 0, 640x360 @ 6,12,24 fps
# Download segment 0, Base Layer
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p-temp/bluesky-III-1080p.seg0-L0.svc
# Download segment 0, temporal EL 1, enhances base layer
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p-temp/bluesky-III-1080p.seg0-L1.svc
# Download segment 0, temporal EL 2, enhances EL 1
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p-temp/bluesky-III-1080p.seg0-L2.svc
# Download segment 0, 1280x720 @ 6,12,24 fps
# Download segment 0, EL 16 (enhances the base layer)
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p-temp/bluesky-III-1080p.seg0-L16.svc
# Download segment 0, EL 17 (enhances EL 16 and EL 1)
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p-temp/bluesky-III-1080p.seg0-L17.svc
# Download segment 0, EL 18 (enhances EL 17 and EL 2)
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p-temp/bluesky-III-1080p.seg0-L18.svc
# Download segment 0, 1920x1080 @ 6,12,24 fps
# Download segment 0, EL 32 (enhances EL 16)
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p-temp/bluesky-III-1080p.seg0-L32.svc
# Download segment 0, EL 33 (enhances EL 17 and EL 32)
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p-temp/bluesky-III-1080p.seg0-L33.svc
# Download segment 0, EL 34 (enhances EL 18 and EL 33)
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p-temp/bluesky-III-1080p.seg0-L34.svc
# Download segment 0, 1920x1080 @ 6,12,24 fps, better quality
# Download segment 0, EL 48 (enhances EL 32)
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p-temp/bluesky-III-1080p.seg0-L48.svc
# Download segment 0, EL 49 (enhances EL 33 and EL 48)
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p-temp/bluesky-III-1080p.seg0-L49.svc
# Download segment 0, EL 50 (enhances EL 34 and EL 49)
wget http://concert.itec.aau.at/SVCDataset/dataset/bluesky/III/segs/1080p-temp/bluesky-III-1080p.seg0-L50.svc
# call svc_merge.py and create the yuv files for the segments
# 6 fps (temporal id = 0)
# svc_merge.py for base layer only (6 fps in this case) - 640x360
python svc_merge.py bluesky-III-1080p.seg0-BL.264 bluesky-III-1080p.init.svc bluesky-III-1080p.seg0-L0.svc
H264AVCDecoderLibTestStatic bluesky-III-1080p.seg0-BL.264 bluesky-III-1080p.seg0-BL.yuv
# svc_merge.py for EL 16 (6 fps in this case) - 1280x720
python svc_merge.py bluesky-III-1080p.seg0-EL16.264 bluesky-III-1080p.init.svc bluesky-III-1080p.seg0-L0.svc bluesky-III-1080p.seg0-L16.svc
H264AVCDecoderLibTestStatic bluesky-III-1080p.seg0-EL16.264 bluesky-III-1080p.seg0-EL16.yuv
# svc_merge.py for EL 32 (6 fps in this case) - 1920x1080
python svc_merge.py bluesky-III-1080p.seg0-EL32.264 bluesky-III-1080p.init.svc bluesky-III-1080p.seg0-L0.svc bluesky-III-1080p.seg0-L16.svc bluesky-III-1080p.seg0-L32.svc
H264AVCDecoderLibTestStatic bluesky-III-1080p.seg0-EL32.264 bluesky-III-1080p.seg0-EL32.yuv
# svc_merge.py for EL 48 (6 fps in this case) - 1920x1080 - high quality
python svc_merge.py bluesky-III-1080p.seg0-EL48.264 bluesky-III-1080p.init.svc bluesky-III-1080p.seg0-L0.svc bluesky-III-1080p.seg0-L16.svc bluesky-III-1080p.seg0-L32.svc bluesky-III-1080p.seg0-L48.svc
H264AVCDecoderLibTestStatic bluesky-III-1080p.seg0-EL48.264 bluesky-III-1080p.seg0-EL48.yuv
# 12 fps (temporal id = 1), depends on 6 fps (temporal id = 0)
# TODO
# 24 fps (temporal id = 2), depends on 12 fps (temporal id = 1) and 6 fps (temporal id = 0)
# TODO
# use mplayer to playback the three yuv files
# 640x360 @ 6,12,24 fps
mplayer -demuxer rawvideo -rawvideo w=640:h=360:format=i420:fps=6 bluesky-III-1080p.seg0-BL.yuv -loop 0
mplayer -demuxer rawvideo -rawvideo w=640:h=360:format=i420:fps=12 bluesky-III-1080p.seg0-EL1.yuv -loop 0
mplayer -demuxer rawvideo -rawvideo w=640:h=360:format=i420:fps=24 bluesky-III-1080p.seg0-EL2.yuv -loop 0
# 1280x720 @ 6,12,24 fps
mplayer -demuxer rawvideo -rawvideo w=1280:h=720:format=i420:fps=6 bluesky-III-1080p.seg0-EL16.yuv -loop 0
mplayer -demuxer rawvideo -rawvideo w=1280:h=720:format=i420:fps=12 bluesky-III-1080p.seg0-EL17.yuv -loop 0
mplayer -demuxer rawvideo -rawvideo w=1280:h=720:format=i420:fps=24 bluesky-III-1080p.seg0-EL18.yuv -loop 0
# 1920x1080 @ 6,12,24 fps
mplayer -demuxer rawvideo -rawvideo w=1920:h=1080:format=i420:fps=6 bluesky-III-1080p.seg0-EL32.yuv -loop 0
mplayer -demuxer rawvideo -rawvideo w=1920:h=1080:format=i420:fps=12 bluesky-III-1080p.seg0-EL33.yuv -loop 0
mplayer -demuxer rawvideo -rawvideo w=1920:h=1080:format=i420:fps=24 bluesky-III-1080p.seg0-EL34.yuv -loop 0
# 1920x1080 @ 6,12,24 fps - high quality
mplayer -demuxer rawvideo -rawvideo w=1920:h=1080:format=i420:fps=6 bluesky-III-1080p.seg0-EL48.yuv -loop 0
mplayer -demuxer rawvideo -rawvideo w=1920:h=1080:format=i420:fps=12 bluesky-III-1080p.seg0-EL49.yuv -loop 0
mplayer -demuxer rawvideo -rawvideo w=1920:h=1080:format=i420:fps=24 bluesky-III-1080p.seg0-EL50.yuv -loop 0
# remove the files we created
rm *.yuv
rm *.svc
rm *.264
cd ..
This section details on how to parse the MPD using libdash, assuming we are in the main directory (DASH-SVC-Toolchain) again.
parseMPD uses libdash to parse the MPD files of our dataset, and prints all segments and layers. This tool can be used for testing our dataset, e.g., downloading a full set of segments and layers, and then decoding them using the tools described in the Decoding section.
LIBDASHPATH=$(pwd)/libdash/libdash/build/bin
LIBDASH=$LIBDASHPATH/libdash.so
cd parseMPD
make
# example 1: Print segments of any MPD file (in this case: http://concert.itec.aau.at/SVCDataset/dataset/mpd/factory-I-720p.mpd)
LD_LIBRARY_PATH=$LIBDASHPATH/ ./parseMPD
# example 2: Print segments of a specified MPD file
LD_LIBRARY_PATH=$LIBDASHPATH/ ./parseMPD http://concert.itec.aau.at/SVCDataset/dataset/mpd/rushhour-I-360p.mpd
# example 3: Print only a specific layer of the specified MPD file
LD_LIBRARY_PATH=$LIBDASHPATH/ ./parseMPD http://concert.itec.aau.at/SVCDataset/dataset/mpd/rushhour-I-360p.mpd 0
# Layer 1
LD_LIBRARY_PATH=$LIBDASHPATH/ ./parseMPD http://concert.itec.aau.at/SVCDataset/dataset/mpd/rushhour-I-360p.mpd 1
# Layer 2
LD_LIBRARY_PATH=$LIBDASHPATH/ ./parseMPD http://concert.itec.aau.at/SVCDataset/dataset/mpd/rushhour-I-360p.mpd 2
The output will contain a list of SVC segment files.
Assuming you are in the DASH-SVC-Toolchain directory, follow these steps:
cd demultiplex
# download a raw svc sequence from the dataset
wget http://concert.itec.aau.at/SVCDataset/svcseqs/IV/bluesky-IV.264
# analyze this sequence
python demultiplex.py -a bluesky-IV.264 48 > analyze_normal.txt
# analyze this sequence with temporal scalability added
python demultiplex.py -a bluesky-IV.264 48 -t 3 > analyze_temporal.txt
# remove the downloaded .264 file
rm bluesky-IV.264
Assuming you are in the DASH-SVC-Toolchain directory, follow these steps:
cd demultiplex
# download a raw svc sequence from the dataset
wget http://concert.itec.aau.at/SVCDataset/svcseqs/IV/bluesky-IV.264
# demultiplex - 24 frames per second, 48 frames per segment, skip 0 frames at the beginning (skip frames only for JSVM), baseURL = http://...
python demultiplex.py bluesky-IV.264 48 dataset/bluesky/IV/segs/1080p/ 24 0 http://concert.itec.aau.at/SVCDataset/ > analyze_normal.txt
Assuming you are in the DASH-SVC-Toolchain directory, follow these steps:
cd demultiplex
# download a raw svc sequence from the dataset
wget http://concert.itec.aau.at/SVCDataset/svcseqs/IV/bluesky-IV.264
# demultiplex - 24 frames per second, 48 frames per segment, skip 0 frames at the beginning (skip frames only for JSVM), baseURL = http://...
# temporal scalability: 3 layers (24 fps, 12 fps, 6 fps)
python demultiplex.py -t 3 bluesky-IV.264 48 dataset/bluesky/IV/segs/1080p/ 24 0 http://concert.itec.aau.at/SVCDataset/ > analyze_normal.txt