Tehnički glasnik, Vol. 20 No. 4, 2026.
Prethodno priopćenje
https://doi.org/10.31803/tg-20250516203854
Aeye: Real-Time Age, Gender and Emotion Detection in Browser
Mihael Holi
; Juraj Dobrila University of Pula, Faculty of Engineering, Alda Negrija 6, 52100 Pula, Croatia
*
Sven Maričić
; 1) Juraj Dobrila University of Pula, Faculty of Engineering, Alda Negrija 6, 52100 Pula, Croatia / 2) University of Rijeka, Faculty of Medicine, Center for Biomodeling and Innovations in Medicine, Br. Branchetta 20, 51000 Rijeka, Croatia
Irena Jurjević
; King's College London, Institute of Psychiatry, Psychology & Neuroscience, Newcomme Street, London, SE1 1UL, The United Kingdom
Adriana Prunk Drmić
; Thalassotherapia Opatija, Maršala Tita 188/I, 51410 Opatija, Croatia
Antonija Ružić Baršić
; Thalassotherapia Opatija, Maršala Tita 188/I, 51410 Opatija, Croatia
* Dopisni autor.
Sažetak
This paper presents Aeye, a lightweight, browser-based application capable of real-time detection of human age, gender and emotional state using artificial intelligence. Unlike many existing AI systems, Aeye performs all computation on the client side, preserving user privacy and enabling full offline functionality. The application utilizes face-api.js and Tiny Face Detector models, achieving fast and efficient performance even on low-resource devices. Its modular design supports dynamic interface controls and responsive operation across platforms. The scientific contribution lies in the integration of asynchronous model loading, real-time GPU-accelerated inference using TensorFlow.js, and deployment as a progressive web app (PWA). These innovations make Aeye suitable for educational, experimental, and field use cases where connectivity and infrastructure are limited. The paper outlines the system architecture, implementation, and evaluation results, and discusses potential extensions such as multi-face detection and integration with emotion-driven user interfaces.
Ključne riječi
age estimation; emotion recognition; face detection; gender classification; real-time AI
Hrčak ID:
351737
URI
Datum izdavanja:
15.12.2026.
Posjeta: 0 *