Documentation

Table of contents

• a. Prerequisites
• B. Installation
• C. How to run
• D. Handle Not Supported CPU
• E. Run App As A Service

A. Prerequisites

• Node.js v16.17.0
• NPM v8.15.0
• Git LFS
• Python 2.7 or newer

B. Installation

1. Clone this repo
2. Change to repository directory
3. Install all required package npm i

C. How to run

• node app.js

If app crash and returned "Illegal Insctructions (Core Dumped)", then continue on section D.

Note: If Program running well till showing "Server started on port" then continue to section E

D. Handle Not Supported CPU (Linux Only)

Prebuild tfjs-node shared object comes with optimization for avx2 computation, to check if cpu not supported run this command

grep flags -m1 /proc/cpuinfo | cut -d ":" -f 2 | tr '[:upper:]' '[:lower:]' | { read FLAGS; OPT="-march=native"; for flag in $FLAGS; do case "$flag" in "sse4_1" | "sse4_2" | "ssse3" | "fma" | "cx16" | "popcnt" | "avx" | "avx2") OPT+=" -m$flag";; esac; done; MODOPT=${OPT//_/\.}; echo "$MODOPT"; }

If there is no output like "-mavx2", run this command to used precompiled shared object without avx2 optimization

cp build/prebuild-ubuntu.tar.gz node_modules/@tensorflow/tfjs-node/deps
tar -xf node_modules/@tensorflow/tfjs-node/deps/prebuild-ubuntu.tar.gz

Then try running again

E. Run App As A Service

If run app like section C, it will stopped if ssh connection closed or server started, so to handle that using a package named pm2.

1. Run npm i -g pm2
2. pm2 startup
3. pm2 startup systems
4. systemctl status pm2-root.service
5. cd/<app-directory>
6. pm2 start app.js -n face-api
7. pm2 save
8. Verify pm2 status

Reference: https://www.tecmint.com/enable-pm2-to-auto-start-node-js-app/