VideoFramesFaceRecognition

Video Frames Face Recognition Samples

Demo: http://aka.ms/amcdemo_videoface

1. Preparation

1-1. Configurations for Azure Services

You have to create the following Azure services accounts and configure the files for each service:

Azure Services	Config file
Azure Media Services	ams.conf
Azure Storage (x)	storage.conf
Cognitive Services (faceapi)	cognitive.conf
(x) - Azure Storage account here has to be the one that is attached to Azure Media Services account here.

1-2. Create faceapi person group

./cogutils/create_persongroup.sh <personGroupId> <groupName>

For example, suppose personGroupId is 'build2016keynote', and groupName is 'DEMO Build 2016 Keynote', the command is like this:

./cogutils/create_persongroup.sh build2016keynote 'DEMO Build 2016 Keynote'

1-3. Create peopleinput.csv

peopleinput.csv is an initial file that need to be created at under 'BATCH_WORK_DIR' directory. The format of peopleinput.csv is

Face Image URL	Face Name	Master Flag (0/1)
http://imgserver.com/face1.jpg	facename1	1
http://imgserver.com/face2.jpg	facename2	1
...	...	1
http://imgserver.com/faceN.jpg	facenameN	0

See sample peopleinput.csv

1-4. Setup Azure Media Processors Modules

Please install maven and git if not yet installed on your environment

# for Ubuntu,Debian
sudo apt-get install git
sudo apt-get install maven

# for CentOS,Fedora,Oracle Linux,Red Hat Enterprise Linux
sudo yum install git
sudo yum install maven

Also you need Java compiler, so please install JDK if not yet installed on your environment

# for Ubuntu,Debian
sudo apt-get install openjdk-7-jdk   (JDK7)
sudo apt-get install openjdk-8-jdk   (JDK8)

# for CentOS,Fedora,Oracle Linux,Red Hat Enterprise Linux
sudo yum install java-1.7.0-openjdk  (JDK7)
sudo yum install java-1.8.0-openjdk  (JDK8)

Finall Check out github repo and build the module in order to make it ready for batch execution

cd mediaprocessors
./setup.sh

2. Batch execution

run-batch command Usages

usage: ./run-batch [WORKFLOW] [BATCH_NAME] [BATCH_WORK_DIR] [PARAMS]
This program generates webvtt file from your video and people master 
that you register leveraging Azure Media Services and Cognitive Services

WORKFLOW:  ALL|REGIST_PEOPLE|GEN_FRAME|IDENTIFY
PARAMS for each WORKFLOW:
  REGIST_PEOPLE => [PERSON_GROUP_ID]
    ex)  ./run_batch REGIST_PEOPLE mybatch /mybatch-work-dir my_person_group_id
  GEN_FRAME => [VIDEO_FILE]
    ex)  ./run_batch GEN_FRAME mybatch /mybatch-work-dir /myvideo-path/myvideo.mp4
  IDENTIFY => [PERSON_GROUP_ID]
    ex)  ./run_batch IDENTIFY mybatch /mybatch-work-dir my_person_group_id
  ALL => [PERSON_GROUP_ID] [VIDEO_FILE]
    ex)  ./run_batch ALL mybatch /mybatch-work-dir my_person_group_id /myvideo-path/myvideo.mp4

Example1: Run all workflows

./run-batch ALL build2016keynote ../demo/build2016keynote demo_build2016keynote

Example2: Run each workflow one by one

# regist people faces example
./run-batch REGIST_PEOPLE nogizaka46 demo_build2016keynote ../demo/build2016keynote

# generate frames
./run-batch GEN_FRAME build2016keynote ../demo/build2016keynote ../demo/build2016keynote/videos/build2016keynote.mp4

# identify
./run-batch IDENTIFY build2016keynote ../demo/build2016keynote demo_build2016keynote

Contributing

Bug reports and pull requests are welcome on GitHub at https://github.com/AzureMediaCognitiveDemos/VideoFramesFaceRecognition-Python.