FaceNet: A Unified Embedding for Face Recognition and Clustering
Transcription
FaceNet: A Unified Embedding for Face Recognition and Clustering
Introduction Algorithm Results References FaceNet: A Unified Embedding for Face Recognition and Clustering Felipe Bombardelli 2015 Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri Introduction Algorithm Results References Sumário 1 Introduction 2 Algorithm Detection Alignment Description Classification 3 Results 4 References Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri Introduction Algorithm Results References Introduction Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri Introduction Algorithm Results References Introduction Detection Detect faces with a pre-trained models from dlib or OpenCV. Alignment Transform the face for the neural network. This repository uses dlib’s real-time pose estimation with OpenCV’s affine transformation to try to make the eyes and nose appear in the same location on each image. Description Use a deep neural network to represent (or embed) the face on a 128-dimensional unit hypersphere. The embedding is a generic representation for anybody’s face. assification Apply your favorite clustering or classification techniques to the features to complete your recognition task. Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri Introduction Algorithm Results References Detection Alignment Description Classification Sumário 1 Introduction 2 Algorithm Detection Alignment Description Classification 3 Results 4 References Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri Introduction Algorithm Results References Detection Alignment Description Classification Sumário 1 Introduction 2 Algorithm Detection Alignment Description Classification 3 Results 4 References Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri Introduction Algorithm Results References Detection Alignment Description Classification Face Alignment Face alignment can be solved with a cascade of regression functions. S (t+1) = S (t) + rt (I , S (t) ) (1) Figure: Regression in step by step Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri Introduction Algorithm Results References Detection Alignment Description Classification Face Alignment Figure: Face with shape predicted by algorithm descripted in One Millisecond Face Alignment with an Ensemble of Regression Trees Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri Introduction Algorithm Results References Detection Alignment Description Classification Facebook’s Algorithm Figure: Alignment 3D to the face in the Facebook’s algorithm Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri Introduction Algorithm Results References Detection Alignment Description Classification Sumário 1 Introduction 2 Algorithm Detection Alignment Description Classification 3 Results 4 References Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri Introduction Algorithm Results References Detection Alignment Description Classification Convolutional Neural Networks (CNN) Figure: DeepFace architecture Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri Introduction Algorithm Results References Detection Alignment Description Classification Convolutional Neural Networks (CNN) Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri Introduction Algorithm Results References Detection Alignment Description Classification Convolutional Neural Networks (CNN) Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri Introduction Algorithm Results References Detection Alignment Description Classification Sumário 1 Introduction 2 Algorithm Detection Alignment Description Classification 3 Results 4 References Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri Introduction Algorithm Results References Detection Alignment Description Classification Classification t-SNE is a dimensionality reduction technique that can be used to visualize the 128-dimensional features OpenFace produces Figure: The following shows the visualization of the three people in the training and testing dataset Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri Introduction Algorithm Results References Results Figure: Classification performance in ILSVRC 2014 Classification Challenge Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri Introduction Algorithm Results References References FaceNet: A Unified Embedding for Face Recognition and Clustering Going deeper with convolutions DeepFace: Closing the Gap to Human-Level Performance in Face Verication One Millisecond Face Alignment with an Ensemble of Regression Trees Network in Network Felipe Bombardelli FaceNet: A Unified Embedding for Face Recognition and Clusteri