In car theft security system by face detection
[ X ]
Tarih
2023
Yazarlar
Dergi Başlığı
Dergi ISSN
Cilt Başlığı
Yayıncı
Altınbaş Üniversitesi / Lisansüstü Eğitim Enstitüsü
Erişim Hakkı
info:eu-repo/semantics/closedAccess
Özet
This thesis presents an innovative approach to Optimize Neural Networks for the sake of
preventing cars thefts by utilizing biologic identification by deep machine learning. The
proposed optimization technique leveraging the power of Yogi algorithm [43] adjusted with
Nesterov Accelerated Gradient (Nesterov Momentum) [48] and weights averaging [42]. This
approach can potentially enhance the performance leading to improved generalization,
reduced overfitting, faster and enhanced convergence.
The model employs a neural network architecture consist input layer (120, 120, 3) followed
by hidden layers and followed by two prediction heads each one start with a Global Max
Pooling with Rectified Linear Units (ReLU) and Dense layer with Sigmoid functions to
perform facial Detection.
A Siamese Network consists of three subnetworks trained to preform facial Recognition, this
neural network uses a triple loss method, consists of three major components
( Embedding layer have three inputs each one followed by a neural network similar to
ResNet50 architecture and output of Dense layer 256 units as a feature vector for the that
input, A Distance layer subtracting the three Embeddings, And a Classification layer
calculated the triplet loss function.
Experiment held using Google Colab and the performance is evaluated through metrics such
as regression losses, classification losses and IoU for the face detection neural network. And
Loss, Accuracy and other positives and negatives calculations that evaluate the Siamese
Neural network.The results demonstrate the efficacy of the proposed method in terms regression and
classifications tasks in Deep neural networks, which can be later used for providing a novel
reliable solution for optimizing neural networks.
Açıklama
Anahtar Kelimeler
Deep Machine Learning, Optimization, Yogi Algorithm, Nesterov Momentum, Siamese Neural Networks, Triplet Loss, Face Detection
Kaynak
WoS Q Değeri
Scopus Q Değeri
Cilt
Sayı
Künye
Altaher, R. H. R. (2023). In car theft security system by face detection. (Yayınlanmamış yüksek lisans tezi). Altınbaş Üniversitesi, Lisansüstü Eğitim Enstitüsü, İstanbul.
