Bash scripts to use ncap-livestream

[curibe]

Bash script to run webrtc and rtsp in neurocaas pipeline

Bash scripts are created to run webrtc and rtsp on the ec2 server deployed by the neurocaas pipeline.

Test locally

The test is performed on a local machine, in order to validate the operation of each script:

• Download of the 'loaded' data to the s3 folder. This step emulates the process of uploading the files to the s3 bucket.
• Execute each server.

This requires to have installed the repos containing the solution of webrtc and rtsp (in the rtmp-livestream branch).

The data used are:

• DLCLive model: DLC_Dog_resnet_50_iteration-0_shuffle-0.tar.gz
• Configuration file required by DLCLive: dlc_config,json
• Yaml configuration file to run RTSP: config_rtsp.yml
• Configuration file read by the webrtc server. It is similar to the one used by dlclive: dlc_config,json

Each of these configuration files are located in the repo of each solution.

WEBRTC

To test webrtc locally, we proceeded as follows:

• Create a temporal folder, s3 folder to save the pipeline process

mkdir tmp
mkdir -p s3/results

• Initialize the job

neurocaas-contrib workflow initialize-job -p ./tmp

• Register the dataset, config file and result path. You need to have the data model downloaded

neurocaas-contrib workflow register-dataset -l ./DLC_Dog_resnet_50_iteration-0_shuffle-0.tar.gz
neurocaas-contrib workflow register-config -l ./dlc_config.json
neurocaas-contrib workflow register-resultpath -l s3/results

• Change permission to script

chmod 700 run_webrtc_local.sh

• Run the script locally

neurocaas-contrib workflow log-command-local -c ./run_webrtc_local.sh

RTSP

To test rtsp locally, we proceeded as follows:

• Create a temporal folder, s3 folder to save the pipeline process

mkdir tmp
mkdir -p s3/results

• Initialize the job

neurocaas-contrib workflow initialize-job -p ./tmp

• Register the dataset, config file and result path. You need to have the data model downloaded.
  neurocaas_contrib only allows to register one data file and one configuration file. In this case, we need to manage a configuration file that allows to control the input parameters of the server script, as well as another configuration file that reads the server to manage other parameters. To implement this, the bash script that runs the server receives:

  • A yaml file as a configuration file (config_rtsp.yml):

  send: video
  cfg_filename: dlc_config.json
  path_cfg_file: Documents/DeepLabCut-live-GUI/config/
  model_file: DLC_Dog_resnet_50_iteration-0_shuffle-0.tar.gz

  • A zip file (rtsp_input.zip) with the model data (DLC_Dog_resnet_50_iteration-0_shuffle-0.tar.gz) and the configuration file it requires (dlc_config.json).

To register the files, we proceed in the same way

neurocaas-contrib workflow register-dataset -l rtsp_input.zip
neurocaas-contrib workflow register-config -l config_rtsp.yml
neurocaas-contrib workflow register-resultpath -l s3/results

• Change permission to script

chmod 700 run_rtsp_zmq_local.sh

• Run the script locally

neurocaas-contrib workflow log-command-local -c ./run_rtsp_zmq_local.sh

NOTE :
The difference between the files run_*_cli.sh (committed) and run_*_local.sh (mentioned here) is the path indicated in the userhome variable