A partir de cette page vous pouvez :
Retourner au premier écran avec les étagères virtuelles... |
Résultat de la recherche
1 résultat(s) recherche sur le mot-clé 'Computer vision, video-based face recognition, face tracking, Kalman Filter, Particle Filter, face recognition system, triplet loss, CNN.'
Affiner la recherche Faire une suggestion
State of the Art on Deep Learning Video-Based Face Recognition and Tracking Systems / LISSAOUI El Yazid
Titre : State of the Art on Deep Learning Video-Based Face Recognition and Tracking Systems Type de document : projet fin études Auteurs : LISSAOUI El Yazid, Auteur Langues : Français (fre) Catégories : BIG DATA Mots-clés : Computer vision, video-based face recognition, face tracking, Kalman Filter, Particle Filter, face recognition system, triplet loss, CNN. Index. décimale : mast 262/19 Résumé : Nowadays, automatic face recognition systems are a prominent subject of research in the field of computer vision for their multiple and increasing applications; they gained more attention after the introduction of deep learning models that proved to be much more versatile and performant in the form of encoding-oriented CNNs. Nonetheless, video-based face recognition and tracking systems revealed to be much more challenging than still image operations; unconstrained environments, large volume of data and the spatial-temporality nature of videos push toward the need of developing robust systems that can overcome the mentioned challenges and aim, potentially, for real-time accurate performance. In that regard, research centered its attention toward this problematic and up until recently, dozens of papers have been published that deal with video-based recognition specifically.
This thesis presents the recent state-of-the-art deep learning techniques and models for video-based face recognition as well as methods for face tracking. The motivation behind this work is to find the potential building blocks to construct an automatic and autonomous system for video-based recognition for possibly real time applications. We will explore the Triplet Loss model and autoencoders for face recognition and review the most recent architectures used to achieve high performance. Moreover, the tracking problem will also be discussed through Kalman Filter and Particle filter and discuss their compatibility with the previously mentioned recognition.
State of the Art on Deep Learning Video-Based Face Recognition and Tracking Systems [projet fin études] / LISSAOUI El Yazid, Auteur . - [s.d.].
Langues : Français (fre)
Catégories : BIG DATA Mots-clés : Computer vision, video-based face recognition, face tracking, Kalman Filter, Particle Filter, face recognition system, triplet loss, CNN. Index. décimale : mast 262/19 Résumé : Nowadays, automatic face recognition systems are a prominent subject of research in the field of computer vision for their multiple and increasing applications; they gained more attention after the introduction of deep learning models that proved to be much more versatile and performant in the form of encoding-oriented CNNs. Nonetheless, video-based face recognition and tracking systems revealed to be much more challenging than still image operations; unconstrained environments, large volume of data and the spatial-temporality nature of videos push toward the need of developing robust systems that can overcome the mentioned challenges and aim, potentially, for real-time accurate performance. In that regard, research centered its attention toward this problematic and up until recently, dozens of papers have been published that deal with video-based recognition specifically.
This thesis presents the recent state-of-the-art deep learning techniques and models for video-based face recognition as well as methods for face tracking. The motivation behind this work is to find the potential building blocks to construct an automatic and autonomous system for video-based recognition for possibly real time applications. We will explore the Triplet Loss model and autoencoders for face recognition and review the most recent architectures used to achieve high performance. Moreover, the tracking problem will also be discussed through Kalman Filter and Particle filter and discuss their compatibility with the previously mentioned recognition.
Réservation
Réserver ce document
Exemplaires
Code barre Cote Support Localisation Section Disponibilité mast 262/19 mast 262/19 LIS Texte imprimé Unité des masters Mast/19 Disponible