2D Face SDK

1. Deep adaptation to Android system features

System compatibility: Compatible with mainstream Android versions (7.0-14.0), and customized systems from various manufacturers including HarmonyOS compatibility mode, MIUI, EMUI, etc., effectively addressing fragmentation issues;

Hardware adaptation: Supports front and rear cameras on Android devices, as well as NPU/CPU/GPU heterogeneous computing. Hardware acceleration can be utilized to enhance algorithm efficiency;

Permission and Compliance: Adheres to Android privacy standards, accesses cameras and storage only after user authorization, and complies with Android 10+ rules including scoped storage and background camera restrictions;

Lightweight deployment: Supports APK splitting and on-demand loading, with the core algorithm library size controlled between 5-20MB to avoid excessive storage usage on mobile devices.

2. Characteristics of core algorithm capabilities

Mobile lightweighting: The algorithm model has been pruned and quantized (e.g.,INT8 quantization), enabling smooth operation on low-end Android phones, with single frame face detection time ≤ 20ms;

Live detection capability: Supports silent liveness (no user cooperation, identified via facial texture and micro-movements) and action liveness (eye blinking, head shaking, mouth opening ), and prevents photo/video spoofing attacks;

Scenario robustness: Adapts to complex scenarios on mobile devices (backlight, low light, phone shaking, various angles), enabling stable detection even for non-frontal faces, with a false detection rate＜0.1%;

Strong real-time performance: Supports real-time camera stream processing, enabling "second-level" face unlocking and identity verification, meeting the interactive experience requirements of mobile users.

3.Development and integration features

Quick integration: With a complete demo project (incuding layout files, permission

configurations and usage examples), integration and debugging can be completed in 1-2days;

Offline operationA: The core functions (detection,recognition,liveness) do not require networking

and only call the network when cloud comparison is needed, ensuring the availability of weak

network scenarios.

1.Technical development aspects

Solving the problem of difficult development of mobile facial algorithms:

No need to master complex technologies such as mobile model optimization, hardware adaptation, or Android permission management.

Core functions can be implemented via API calls, effectively lowering the development threshold.

Solving the problem of "difficult adaptation to Android fragmentation":

The SDK has been pre-adapted to Android devices of different brands, versions, and hardware, avoiding the need for developers to debug compatibility issues one by one.

Addressing the issue of balancing algorithm performance and power consumption: optimizing computing power scheduling based on the battery characteristics of Android devices, reducing power consumption while ensuring recognition speed, and avoiding phone overheating.

2.Business application aspects

Identity verification scenarios: addressing facial verification needs for app login, payment, and real-name authentication, replacing passwords and SMS verification codes to enhance security and convenience.

Life service scenarios: solve face verification for food delivery / express locker pickup, hotel check-in, and scenic spot ticket checking, with no need to carry physical ID, improving service efficiency;

Intelligent terminal scenarios: solving "face unlock / clock-in" issues for Android smart door locks, tablet attendance machines and self-service terminals, adapting to the portable nature of mobile devices;

Social entertainment scenarios: solving issues of facial effects, age/gender analysis and expression recognition in short video & beauty apps, enriching product interaction experience;

3.Security and experience aspects

Addressing the issue of "facial forgery attacks": using liveness detection technology to prevent fraudulent methods such as photos, videos, and 3D masks, and ensuring the security of identity verification;

Addressing the issue of "weak/no network usage on mobile devices": offline operation capability ensures that facial functions still work normally in weak or no-network scenarios such as subways and mountainous areas;

Addressing the issue of poor user experience: optimizing mobile interaction logic (such as guiding users to face the camera or indicating insufficient lighting), reducing verification failure rates, and improving user experience.