International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 09 | Sep 2024
p-ISSN: 2395-0072
www.irjet.net
AI-Driven Video Face Recognition: Leveraging Tolerance Optimization For Enhanced Precision Hrishikesh. R. Tavar1, Parth Barse2, Vedant Badre3 1Bachelor of Engineering, Department of Information Technology, AISSMS Institute of Information Technology,
Pune, Maharashtra, India.
2Students, Department of Artificial intelligence & Data Science, AISSMS Institute of Information Technology, Pune,
Maharashtra, India. -----------------, ----------------------------------------------------***---------------------------------------------------------------------
Abstract - An This paper delves into the utilization of
real-time comparison for duplication detection (Jones & Brown, 2019).
Artificial Intelligence (AI) within video face recognition, specifically focusing on enhancing accuracy through the optimization of tolerance parameters. It aims to refine the precision and dependability of face recognition systems by fine-tuning the sensitivity of the classification model. Utilizing Python's face_recognition and cv2 libraries, the study establishes a methodological framework for video processing and facial feature extraction. The methodology encompasses defining input and output directories, initializing variables, and iteratively comparing facial encodings to detect similarities. A crucial aspect involves adjusting the tolerance hyper-parameter to determine face matches. By systematically varying the tolerance level, the experiment aims to identify an optimal setting that balances sensitivity and specificity in face recognition. Results show significant accuracy improvements with the optimized tolerance parameter, minimizing false positives and negatives and enhancing precision and recall rates. The study emphasizes the importance of parameter fine-tuning in AI-driven systems, offering recommendations for practitioners. It contributes to AI-based face recognition technology by introducing a systematic approach to tolerance optimization, enhancing accuracy and reliability in identifying faces from videos. These findings underscore the significance of parameter tuning in AI methodologies and provide insights for future research and practical applications in computer vision.
Currently, there is a notable absence of comprehensive tools capable of efficiently handling vast sets of video input data and conducting real-time comparisons for video duplication (Chen et al., 2021). This gap in the existing technological landscape underscores the urgent need for innovative approaches to address this pressing issue. Motivated by this need, our research endeavors were spurred by a remarkable aspiration: to secure a Guinness World Record for Largest online video album of people saying the same sentence (GHE BHARARI, RAHUL KULKARNI, NEELAM EDLABADKAR, 2024). This ambitious project not only served as a testament to our commitment to pushing the boundaries of technological innovation but also highlighted the paramount importance of developing robust tools for auditing and ensuring the authenticity of extensive video datasets. The primary objective of our project was to devise a novel algorithm capable of significantly reducing the time complexity associated with comparing various videos in real-time while maintaining exceptionally high accuracy in the results (Lee & Kim, 2022). By harnessing the power of Artificial Intelligence (AI) and advanced video processing techniques, our approach aimed to revolutionize the landscape of video duplication detection.
Key Words: Face Recognition, Artificial Intelligence (AI), Video Processing, Tolerance Optimization, Computer Vision
In this introduction, we provide a critical analysis of existing solutions known from scientific literature, highlighting their shortcomings and limitations in addressing the challenges posed by large-scale video datasets (Wang et al., 2020). Subsequently, we present the scientific novelty and advantages of our proposed approach, emphasizing its potential to redefine the standards of efficiency and accuracy in video authentication and duplication detection.
1.INTRODUCTION This In recent years, the exponential growth of digital content, particularly in the form of videos, has posed unprecedented challenges for content management and authentication (Smith et al., 2020). With the proliferation of user-generated content across online platforms, ensuring the authenticity and non-duplication of large datasets of videos has become increasingly crucial. However, existing tools and solutions are often inadequate in handling the sheer volume of video data and providing
© 2024, IRJET
|
Impact Factor value: 8.315
1.1 Literature Survey The literature survey provides an overview of existing research and methodologies related to video duplication
|
ISO 9001:2008 Certified Journal
|
Page 775