Check if processor specific optimization is turned on during build.

https://api.travis-ci.com/v3/job/201936760/log.txt

1: Metadata:
1: encoder : Lavc58.35.100 libx264
1: Side data:
1: cpb: bitrate max/min/avg: 0/0/0 buffer size: 0 vbv_delay: -1
1: frame= 16 fps=0.0 q=0.0 size= 0kB time=00:00:00.00 bitrate=N/A speed= 0x
frame= 31 fps= 31 q=0.0 size= 0kB time=00:00:00.00 bitrate=N/A speed= 0x
frame= 46 fps= 30 q=0.0 size= 0kB time=00:00:00.00 bitrate=N/A speed= 0x
av_interleaved_write_frame(): Broken pipe
1: [flv @ 0x55f29e254a40] Failed to update header with correct duration.
1: [flv @ 0x55f29e254a40] Failed to update header with correct filesize.
1: Error writing trailer of rtmp://localhost/dash/test: Broken pipe
1: frame= 49 fps= 18 q=29.0 Lsize= 49kB time=-00:00:00.03 bitrate=N/A speed=N/A
1: video:94kB audio:0kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: unknown
1: [libx264 @ 0x55f29e255f00] frame I:1 Avg QP:36.00 size: 48644
1: [libx264 @ 0x55f29e255f00] frame P:48 Avg QP:36.00 size: 47025
1: [libx264 @ 0x55f29e255f00] mb I I16..4: 0.0% 0.0% 100.0%
1: [libx264 @ 0x55f29e255f00] mb P I16..4: 100.0% 0.0% 0.0% P16..4: 0.0% 0.0% 0.0% 0.0% 0.0% skip: 0.0%
1: [libx264 @ 0x55f29e255f00] 8x8 transform intra:0.0%
1: [libx264 @ 0x55f29e255f00] coded y,uvDC,uvAC intra: 100.0% 100.0% 100.0%
1: [libx264 @ 0x55f29e255f00] i16 v,h,dc,p: 0% 0% 95% 5%
1: [libx264 @ 0x55f29e255f00] i4 v,h,dc,ddl,ddr,vr,hd,vl,hu: 3% 3% 34% 12% 10% 11% 10% 9% 9%
1: [libx264 @ 0x55f29e255f00] i8c dc,h,v,p: 93% 0% 0% 7%
1: [libx264 @ 0x55f29e255f00] Weighted P-Frames: Y:0.0% UV:0.0%
1: [libx264 @ 0x55f29e255f00] kb/s:11293.92
1: Conversion failed!

https://github.com/opencv/dldt/releases/tag/2019_R1.1

ffmpeg+gst+dev packages also include DLDT.
Change it to ffmpeg+gst+dldt+dev ??

Logs shown as below:

Step 21/63 : RUN wget -O - ${MP3LAME_REPO} | tar xz; cd lame-${MP3LAME_VER}; ./configure --prefix="/usr" --libdir=/usr/lib64 --enable-shared --enable-nasm; make -j8; make install DESTDIR=/home/build; make install
---> Running in 4e904c50ec63
--2019-03-06 02:49:20-- https://sourceforge.net/projects/lame/files/lame/3.100/lame-3.100.tar.gz
Resolving sourceforge.net (sourceforge.net)... 216.105.38.13
Connecting to sourceforge.net (sourceforge.net)|216.105.38.13|:443... connected.
Unable to establish SSL connection.
gzip: stdin: unexpected end of file
tar: Child returned status 1
tar: Error is not recoverable: exiting now
/bin/sh: line 0: cd: lame-3.100: No such file or directory
/bin/sh: ./configure: No such file or directory
make: *** No targets specified and no makefile found. Stop.
make: *** No rule to make target install'. Stop. make: *** No rule to make target install'. Stop.
The command '/bin/sh -c wget -O - ${MP3LAME_REPO} | tar xz; cd lame-${MP3LAME_VER}; ./configure --prefix="/usr" --libdir=/usr/lib64 --enable-shared --enable-nasm; make -j8; make install DESTDIR=/home/build; make install' returned a non-zero code: 2
Xeon/centos-7.4/dldt+ffmpeg/CMakeFiles/build_xeon_dldt_ffmpeg_cent74.dir/build.make:57: recipe for target 'Xeon/centos-7.4/dldt+ffmpeg/CMakeFiles/build_xeon_dldt_ffmpeg_cent74' failed
make[2]: *** [Xeon/centos-7.4/dldt+ffmpeg/CMakeFiles/build_xeon_dldt_ffmpeg_cent74] Error 2
CMakeFiles/Makefile2:2079: recipe for target 'Xeon/centos-7.4/dldt+ffmpeg/CMakeFiles/build_xeon_dldt_ffmpeg_cent74.dir/all' failed
make[1]: *** [Xeon/centos-7.4/dldt+ffmpeg/CMakeFiles/build_xeon_dldt_ffmpeg_cent74.dir/all] Error 2
Makefile:94: recipe for target 'all' failed
make: *** [all] Error 2
The command "make" exited with 2.

https://api.travis-ci.com/v3/job/201936746/log.txt

4: New clock: GstSystemClock
4: ERROR: from element /GstPipeline:pipeline0/GstDecodeBin:decodebin0/GstVP9Dec:vp9dec0: Failed to decode frame
4: Additional debug info:
4: gstvpxdec.c(696): gst_vpx_dec_handle_frame (): /GstPipeline:pipeline0/GstDecodeBin:decodebin0/GstVP9Dec:vp9dec0:
4: corrupt frame
4: Execution ended after 0:00:00.472492924
4: Setting pipeline to PAUSED ...
4: Setting pipeline to READY ...
4: Setting pipeline to NULL ...
4: Freeing pipeline ...
4/10 Test #4: test_xeon_dldt_gst_centos74_gst_svt ....................***Failed 14.14 sec

Please use conditional macros to differentiate media and analytics images:

Fetch FFmpeg source

ARG FFMPEG_VER=n4.1
ARG FFMPEG_REPO=https://github.com/FFmpeg/FFmpeg/archive/${FFMPEG_VER}.tar.gz
ARG FFMPEG_FLV_PATCH_REPO=https://raw.githubusercontent.com/VCDP/CDN/master/The-RTMP-protocol-extensions-for-H.265-HEVC.patch
ARG FFMPEG_1TN_PATCH_REPO=https://patchwork.ffmpeg.org/patch/11625/raw
ARG FFMPEG_THREAD_PATCH_REPO=https://patchwork.ffmpeg.org/patch/11035/raw
ARG FFMPEG_MA_PATCH_REPO_01=https://raw.githubusercontent.com/VCDP/FFmpeg-patch/master/media-analytics/0001-Intel-inference-engine-detection-filter.patch
ARG FFMPEG_MA_PATCH_REPO_02=https://raw.githubusercontent.com/VCDP/FFmpeg-patch/master/media-analytics/0002-New-filter-to-do-inference-classify.patch
ARG FFMPEG_MA_PATCH_REPO_03=https://raw.githubusercontent.com/VCDP/FFmpeg-patch/master/media-analytics/0003-iemetadata-convertor-muxer.patch
ARG FFMPEG_MA_PATCH_REPO_04=https://raw.githubusercontent.com/VCDP/FFmpeg-patch/master/media-analytics/0004-Kafka-protocol-producer.patch
ARG FFMPEG_MA_PATCH_REPO_05=https://raw.githubusercontent.com/VCDP/FFmpeg-patch/master/media-analytics/0005-Support-object-detection-and-featured-face-identific.patch
ARG FFMPEG_MA_PATCH_REPO_06=https://raw.githubusercontent.com/VCDP/FFmpeg-patch/master/media-analytics/0006-Send-metadata-in-a-packet-and-refine-the-json-format.patch
ARG FFMPEG_MA_PATCH_REPO_07=https://raw.githubusercontent.com/VCDP/FFmpeg-patch/master/media-analytics/0007-Refine-features-of-IE-filters.patch
ARG FFMPEG_MA_PATCH_REPO_08=https://raw.githubusercontent.com/VCDP/FFmpeg-patch/master/media-analytics/0008-fixed-extra-comma-in-iemetadata.patch
ARG FFMPEG_MA_PATCH_REPO_09=https://raw.githubusercontent.com/VCDP/FFmpeg-patch/master/media-analytics/0009-add-source-as-option-source-url-calculate-nano-times.patch
ARG FFMPEG_MA_PATCH_REPO_10=https://raw.githubusercontent.com/VCDP/FFmpeg-patch/master/media-analytics/0010-fixed-buffer-overflow-issue-in-iemetadata.patch

RUN DEBIAN_FRONTEND=noninteractive apt-get update && apt-get install -y -q --no-install-recommends zlib1g-dev libssl-dev

RUN wget -O - ${FFMPEG_REPO} | tar xz && mv FFmpeg-${FFMPEG_VER} FFmpeg &&
cd FFmpeg &&
wget -O - ${FFMPEG_FLV_PATCH_REPO} | patch -p1 &&
wget -O - ${FFMPEG_1TN_PATCH_REPO} | patch -p1 &&
wget -O - ${FFMPEG_THREAD_PATCH_REPO} | patch -p1 &&
wget -O - ${FFMPEG_MA_PATCH_REPO_01} | patch -p1 &&
wget -O - ${FFMPEG_MA_PATCH_REPO_02} | patch -p1 &&
wget -O - ${FFMPEG_MA_PATCH_REPO_03} | patch -p1 &&
wget -O - ${FFMPEG_MA_PATCH_REPO_04} | patch -p1 &&
wget -O - ${FFMPEG_MA_PATCH_REPO_05} | patch -p1 &&
wget -O - ${FFMPEG_MA_PATCH_REPO_06} | patch -p1 &&
wget -O - ${FFMPEG_MA_PATCH_REPO_07} | patch -p1 &&
wget -O - ${FFMPEG_MA_PATCH_REPO_08} | patch -p1 &&
wget -O - ${FFMPEG_MA_PATCH_REPO_09} | patch -p1 &&
wget -O - ${FFMPEG_MA_PATCH_REPO_10} | patch -p1;

ffmpeg -i travel6.mp4 -c:v libsvt_av1 -y travel6_av1.mp4
ffprobe -v error -show_streams travel6_av1.mp4

Probing errors into this:
root@6298ae4ea5e2:/home# x
[mov,mp4,m4a,3gp,3g2,mj2 @ 0x55ededd3fd40] Empty AV1 Codec Configuration Box
[mov,mp4,m4a,3gp,3g2,mj2 @ 0x55ededd3fd40] error reading header
travel6_av1.mp4: Invalid data found when processing input

Gstreamer gets built with gst-vaapi, but h264enc test fails with internal data stream error

Verified it works for DAV1D plugin in FFMPEG master branch, and waiting for below dependencies ready for integrating it into Dockerfiles -

1. FFMPEG stable release (v4.2).
2. Existing FFMPEG plugin(i.e. SVT) support FFMPEG v4.2

Dockerfiles v1.0 release is planned by End of March, 2019

Code freeze time: 27, Mar, 2019

Cards can be added to your board to track the progress of issues and pull requests. You can also add note cards, like this one!

Dockerfiles CI build leverage Travis CI cloud service with a limitation of build time should be <50 mins, what's the opportunity to speed up our dockerfiles build and improve it?

that include all SVT codecs.

For example, we are building some dockerfile with dldt (you can see example below) on a PC with 4 cores and not enough RAM this build process will be endless (it will hang).

RUN git clone -b ${DLDT_VER} ${DLDT_REPO}; \
    cd dldt; \
    git submodule init; \
...
    make -j16; \
    rm -rf ../bin/intel64/Release/lib/libgtest*; \

We can take into account number of cores by using nproc then above step will look like this:

RUN git clone -b ${DLDT_VER} ${DLDT_REPO}; \
    cd dldt; \
    git submodule init; \
...
    make -j $(nproc); \
    rm -rf ../bin/intel64/Release/lib/libgtest*; \

The svt-hevc version is not updated.

Following test fails sporadically in travis on random centos ffmpeg images:

/mnt/ffmpeg_svt_hevc.sh: line 4: 9 Segmentation fault (core dumped) ffmpeg -f rawvideo -vcodec rawvideo -s 320x240 -r 30 -pix_fmt yuv420p -i test.yuv -c:v libsvt_hevc -y test.mp4
6/10 Test #6: test_xeon_ffmpeg_centos75_ffmpeg_svt_hevc ...***Failed 1.88 sec

Log from master:
https://api.travis-ci.com/v3/job/199781932/log.txt
Log from pull request:
https://api.travis-ci.com/v3/job/199645510/log.txt

Need documentation/example of how to build and deploy a C++ application using the combination of the development and the deployment images.

To make better structure and easy to understand and find the image, as more and more images joined in Open Visual Cloud Dockerfiles

We need to avoid usage of semicolons.
In case one command finished with failure the RUN won't be marked as failed it just execute next command after semicolon.

For example:

RUN git clone -b ${DLDT_VER} ${DLDT_REPO}; \
    cd dldt; \
    git submodule init; \
    git submodule update --recursive; \
    cd inference-engine; \
    wget -O - ${DLDT_C_API_REPO} | patch -p2; \

If git clone in the above command will fail this RUN won't fail and you may not even notice that something goes wrong, it just execute next commands.

https://github.com/OpenVisualCloud/SVT-AV1/releases

Huge STDOUT for gst dev packge results in big raw file log for travis. Causing it to exceed travis raw log limit.
Figure out:

• limit log
• how to increase log limit on travis if possible

By submitting to dockerhub, plan to redistribute only the “software” layers.
A docker image constructs of multiple layers, with the bottom layer being the OS. The OS image is not redistributed.

Please see this article: https://www.projectatomic.io/blog/2015/08/why-we-dont-let-non-root-users-run-docker-in-centos-fedora-or-rhel/

I think we can put each component into its own stage.
When one component has dependencies on other components we can install those components like in the final image.
It makes us available to reuse stages that were not changed.

For example:

• dldt has dependencies only on CMake and OpenCV hence we install into dldt's stage CMake and OpenCV
• gstreamer has dependencies on CMake and OpenCV - we install into dldt's stage CMake and OpenCV

When we need to update dldt we do not need to rebuild gstreamer, 'cause gstreamer doesn't depend on dldt.

There is a SVT-AV1 folder left under /home in the final image.
Also, SVTAv1UnitTests are in the final image.
Please remove those.

https://gstreamer.freedesktop.org/releases/1.16/

Version 1.16 brings performance improvements and new features.

./shell.sh by default assumes /dev/dri is present. which is not the case always