Overview: LIQA v5 [doc v15]

This page describes the LIQA library

LIQA stands for "Live Image Quality Assurance". It is an AI-powered tool for real-time, no-reference image-quality assessment based on captured video via a mobile device camera.

Technically, LIQA is a JavaScript library with math and UI inside that is designed to be integrated into a company's web page. LIQA requires no server connection for calculations, utilizing only the device's computational capabilities.

LIQA estimates image-quality parameters and interacts with the user through visual feedback. Through this feedback, LIQA guides users toward standardized images with respect to preset quality parameters, which are defined based on Haut.AI’s expertise in facial skin analysis.

Learn how to try LIQA here: Try LIQA: access request!

What's new in LIQA v5?

The 5th version of LIQA brings several new features and some major fixes. The main two features available from 5.0 are auto-capturing of selfie images and prediction of landmarks (facial key points) on resulting selfie image. The fixes change the behaviour of initialization, add new ways for more flexible LIQA loading and improves the collection flow. See more in What's New: release notes

Library features

Features	Description
Use case scenario	Organize repeatable and trustable image collection experiments following predefined quality standards.
Rejection strictness	Fixed for the head position: distance from camera, inclines, rotation. Configurable for illumination: Two settings to exclude jaw/forehead areas from analysis for specific cases Three settings to allow more or fewer illumination imperfections See more in Config
Estimated parameters	Head rotation Face illumination intensity Face illumination uniformness Camera resolution
Visual feedback	Demonstrates illumination map on a face Returns emitters with parameters' status to allow frontend to interpret values See more in Data collection
Image saving formats	PNG (lossless, high weight) JPEG (no compression) JPEG (25% compression) JPEG (50% compression, minimal weight) See more in Config
Recommended camera resolution	1.2Mp - 7Mp
Maximum image resolution	2000 x 2000 pixels
Minimal required mobile device RAM	3Gb