| Titre : | facial recognition using convolutional neural networks |
| Auteurs : | Mohamed Zeghlache, Auteur ; Fatima Zohra Torki, Directeur de thèse |
| Type de document : | Monographie imprimée |
| Editeur : | Biskra [Algérie] : Faculté des Sciences Exactes et des Sciences de la Nature et de la Vie, Université Mohamed Khider, 2019 |
| Format : | 1 vol. (77 p.) / ill. / 29 cm |
| Langues: | Anglais |
| Mots-clés: | facial recognition,convolutional neural networks,deep learn- ing. |
| Résumé : |
Identifying a person from his or her face is an easy task for humans. Is it the same for a machine? This defi?nes the problem of automatic face recognition,which has generated a great deal of research in recent years. Several tech-niques have been developed for face recognition, The purpose of this memoir is to implement a facial recognition application capable of recognizing faces using the convolutional neural network technique. Given the amount of po-tential software (security, social networks, ...) that can be based on this application, it must meet the requirements of speed and robust results. In this sense, the ?rst part of the application is to locate the faces in the test images. The second part of the application deals with the recognition of localized faces using a version of the VggNet architecture which is one of the CNN architectures. The results we have obtained are very encouraging. |
| Sommaire : |
General Introduction 2 1 State of the art 3 I Biometry and facial recognition systems 4 I.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 I.2 Biometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 I.2.1 De?nition of biometry . . . . . . . . . . . . . . . . . . 5 I.2.2 Properties of a biometric modality . . . . . . . . . . . 5 I.2.3 Biometric modalities . . . . . . . . . . . . . . . . . . . 6 I.2.4 Biometric systems . . . . . . . . . . . . . . . . . . . . . 8 I.2.5 Structure of a biometric system . . . . . . . . . . . . . 11 I.2.6 Performance of biometric systems . . . . . . . . . . . . 12 I.3 Facial recognition . . . . . . . . . . . . . . . . . . . . . . . . . 14 I.3.1 Why facial recognition ? . . . . . . . . . . . . . . . . . 14 I.3.2 Facial recognition system . . . . . . . . . . . . . . . . . 14 I.3.2.1 Image acquisition . . . . . . . . . . . . . . . . 15 I.3.2.2 Detection . . . . . . . . . . . . . . . . . . . . 15 I.3.2.3 Preprocessing . . . . . . . . . . . . . . . . . . 15 I.3.2.4 Feature extraction . . . . . . . . . . . . . . . 15 I.3.2.5 Classi?cation . . . . . . . . . . . . . . . . . . 16 I.3.2.6 Learning . . . . . . . . . . . . . . . . . . . . . 16 I.3.2.7 Decision . . . . . . . . . . . . . . . . . . . . . 16 I.3.3 Di?culties of facial recognition . . . . . . . . . . . . . 16 I.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 II Face detection and recognition methods 18 II.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 II.2 Face recognition techniques . . . . . . . . . . . . . . . . . . . 19 II.2.1 Global methods . . . . . . . . . . . . . . . . . . . . . . 19 II.2.2 Local methods . . . . . . . . . . . . . . . . . . . . . . 19 II.2.3 Hybrid methods . . . . . . . . . . . . . . . . . . . . . . 20 II.3 Face detection algorithms . . . . . . . . . . . . . . . . . . . . 20 II.3.1 Viola-Jones (HAAR CASCADE) . . . . . . . . . . . . 20 II.3.2 Histogram of Oriented Gradients "HOG" . . . . . . . . 21 II.4 Face databases . . . . . . . . . . . . . . . . . . . . . . . . . . 23 II.4.1 Labeled faces in the wild (LFW) . . . . . . . . . . . . 23 II.4.2 FERET Database . . . . . . . . . . . . . . . . . . . . . 24 II.4.3 The AR Database . . . . . . . . . . . . . . . . . . . . . 24 II.4.4 ORL Database . . . . . . . . . . . . . . . . . . . . . . 24 II.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 III Neural networks 26 III.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 III.2 Arti?cial neural network . . . . . . . . . . . . . . . . . . . . . 28 III.2.1 De?nition of ANNs . . . . . . . . . . . . . . . . . . . . 28 III.2.2 History and inspiration behind ANNs . . . . . . . . . . 29 III.2.3 Architecture of ANNs . . . . . . . . . . . . . . . . . . 30 III.2.4 Learning paradigms . . . . . . . . . . . . . . . . . . . . 31 III.2.5 Modeling of ANNs . . . . . . . . . . . . . . . . . . . . 32 III.2.6 A few models of ANNs . . . . . . . . . . . . . . . . . . 35 III.3 Convolutional neural network . . . . . . . . . . . . . . . . . . 36 III.3.1 What is and why CNN ? . . . . . . . . . . . . . . . . . 36 III.3.2 Layers in CNN . . . . . . . . . . . . . . . . . . . . . . 36 III.3.2.1 Convolution layer . . . . . . . . . . . . . . . . 37 III.3.2.2 Pooling layer . . . . . . . . . . . . . . . . . . 39 III.3.2.3 Fully connected layer . . . . . . . . . . . . . . 40 III.3.3 CNN architectures . . . . . . . . . . . . . . . . . . . . 41 III.3.3.1 AlexNet (2012) . . . . . . . . . . . . . . . . . 41 III.3.3.2 GoogLeNet/Inception(2014) . . . . . . . . . . 42 III.3.3.3 ResNet(2015) . . . . . . . . . . . . . . . . . . 43 III.3.3.4 VGGNet (2014) . . . . . . . . . . . . . . . . . 44 III.3.4 VggNet CNN Classi?cation . . . . . . . . . . . . . . . 45 III.3.4.1 VggNet Model training . . . . . . . . . . . . . 46 III.3.4.2 VggNet Model testing . . . . . . . . . . . . . 47 III.3.4.3 Non Linearity . . . . . . . . . . . . . . . . . . 48 III.3.4.4 Softmax Function: . . . . . . . . . . . . . . . 50 III.3.4.5 Cross Entropy Loss: . . . . . . . . . . . . . . 50 III.3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . 51 2 Experimentations, Results and Discussions 52 IV Experimentation 53 IV.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 IV.2 Working environment . . . . . . . . . . . . . . . . . . . . . . . 55 IV.3 Required Packages and libraries . . . . . . . . . . . . . . . . . 55 IV.4 Project Structure . . . . . . . . . . . . . . . . . . . . . . . . . 56 IV.5 Our Dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 IV.6 Our CNN Architecture . . . . . . . . . . . . . . . . . . . . . . 58 IV.7 Training our CNN . . . . . . . . . . . . . . . . . . . . . . . . 62 IV.8 Testing our CNN . . . . . . . . . . . . . . . . . . . . . . . . . 65 IV.9 Limitations of our model . . . . . . . . . . . . . . . . . . . . . 71 IV.10 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 General Conclusion 73 |
Disponibilité (1)
| Cote | Support | Localisation | Statut |
|---|---|---|---|
| MINF/434 | Mémoire master | bibliothèque sciences exactes | Consultable |
