Copyright © 2025 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{181400,
        author = {Veeraj R. Humbe},
        title = {Advanced Facial Recognition Using CNN, Clustering Techniques, And Large Language Models With Open-Source AI APIs For Multilingual Translation},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {12},
        number = {1},
        pages = {4667-4672},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=181400},
        abstract = {Facial recognition systems are increasingly becoming integral to modern security, surveillance, and access control frameworks. Their evolution from rudimentary geometric analysis to deep learning-driven pipelines highlights the rapid innovation in this field. Yet, challenges remain in delivering high accuracy across diverse environments, ensuring scalability, and providing linguistic inclusivity and explainability for broader accessibility.
This paper proposes a holistic facial recognition framework built upon state-of-the-art Convolutional Neural Networks (CNNs), unsupervised clustering methods, and transformer-based Large Language Models (LLMs), complemented with real-time translation via open-source AI APIs. Our system not only ensures high facial recognition accuracy but also incorporates intelligent clustering for dynamic identity management, semantic event explanation, and real-time multilingual interaction.
Through the integration of DBSCAN and HDBSCAN clustering, the system efficiently handles unknown or unlabelled identities, allowing adaptive learning over time. The addition of GPT-4 enables contextual understanding of facial recognition outputs and creates human-readable summaries for administrative monitoring. Furthermore, using open-source APIs like Hugging Face's MarianMT and OpenAI’s Whisper, the system supports text and voice translation across over 50 languages, promoting global inclusivity and enhanced accessibility.
Empirical evaluations on real-world datasets such as VGGFace2 demonstrate impressive accuracy, cluster purity, and language translation performance. This framework showcases a future-ready architecture for building intelligent, scalable, multilingual, and explainable facial recognition systems adaptable across sectors like smart cities, education, defences, and healthcare.},
        keywords = {Facial Recognition, Convolutional Neural Networks, Deep Learning, Unsupervised Clustering, Large Language Models, AI Translation, Accessibility, Real-Time Systems, Ethical AI, Open-Source Integration, Smart Surveillance.},
        month = {June},
        }