RetinaFace: Single-stage Dense Face Localisation in the Wild (2019-05-02T00:00:00.000000Z)

TL;DR

A robust single-stage face detector, named RetinaFace, which performs pixel-wise face localisation on various scales of faces by taking advantages of joint extra-supervised and self-super supervised multi-task learning.

Abstract

Though tremendous strides have been made in uncontrolled face detection, accurate and efficient face localisation in the wild remains an open challenge. This paper presents a robust single-stage face detector, named RetinaFace, which performs pixel-wise face localisation on various scales of faces by taking advantages of joint extra-supervised and self-supervised multi-task learning. Specifically, We make contributions in the following five aspects: (1) We manually annotate five facial landmarks on the WIDER FACE dataset and observe significant improvement in hard face detection with the assistance of this extra supervision signal. (2) We further add a self-supervised mesh decoder branch for predicting a pixel-wise 3D shape face information in parallel with the existing supervised branches. (3) On the WIDER FACE hard test set, RetinaFace outperforms the state of the art average precision (AP) by 1.1% (achieving AP equal to 91.4%). (4) On the IJB-C test set, RetinaFace enables state of the art methods (ArcFace) to improve their results in face verification (TAR=89.59% for FAR=1e-6). (5) By employing light-weight backbone networks, RetinaFace can run real-time on a single CPU core for a VGA-resolution image. Extra annotations and code have been made available at: this https URL.

Authors

Jiankang Deng

11 papers

J. Guo

6 papers

S. Zafeiriou

18 papers

TL;DR

Abstract

Authors

References79 items

Dense 3D Face Decoding Over 2500FPS: Joint Texture & Shape Convolutional Mesh Decoders

WIDER Face and Pedestrian Challenge 2018: Methods and Results

Improved Selective Refinement Network for Face Detection

Robust and High Performance Face Detector

Deformable ConvNets V2: More Deformable, Better Results

Learning Better Features for Face Detection with Feature Fusion and Segmentation Supervision

DSFD: Dual Shot Face Detector

Selective Refinement Network for High Performance Face Detection

On Learning 3D Face Morphable Model from In-the-Wild Images

Generating 3D faces using Convolutional Mesh Autoencoders

Joint Pose and Expression Modeling for Facial Expression Recognition

Mean-Variance Loss for Deep Age Estimation from a Face

Unsupervised Training for 3D Morphable Model Regression

Nonlinear 3D Face Morphable Model

PyramidBox: A Context-assisted Single Shot Face Detector

Joint 3D Face Reconstruction and Dense Alignment with Position Map Regression Network

Seeing Small Faces from Robust Anchor's Perspective

Face Detection Using Improved Faster RCNN

DensePose: Dense Human Pose Estimation in the Wild

IARPA Janus Benchmark - C: Face Dataset and Protocol

CosFace: Large Margin Cosine Loss for Deep Face Recognition

ArcFace: Additive Angular Margin Loss for Deep Face Recognition

Feature Agglomeration Networks for Single Stage Face Detection

Face Attention Network: An Effective Face Detector for the Occluded Faces

Wing Loss for Robust Facial Landmark Localisation with Convolutional Neural Networks

S^3FD: Single Shot Scale-Invariant Face Detector

SSH: Single Stage Headless Face Detector

Focal Loss for Dense Object Detection

Recurrent Scale Approximation for Object Detection in CNN

AgeDB: The First Manually Collected, In-the-Wild Age Database

Scale-Aware Face Detection

Face Detection through Scale-Friendly Deep Convolutional Networks

Face R-CNN

SphereFace: Deep Hypersphere Embedding for Face Recognition

MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications

How Far are We from Solving the 2D & 3D Face Alignment Problem? (and a Dataset of 230,000 3D Facial Landmarks)

Mask R-CNN

Deformable Convolutional Networks

YOLO9000: Better, Faster, Stronger

Finding Tiny Faces

Feature Pyramid Networks for Object Detection

DenseReg: Fully Convolutional Dense Shape Regression In-the-Wild

To boost or not to boost? On the limits of boosted trees for object detection

MS-Celeb-1M: A Dataset and Benchmark for Large-Scale Face Recognition

A Unified Multi-scale Deep Convolutional Neural Network for Fast Object Detection

Supervised Transformer Network for Efficient Face Detection

Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering

Unconstrained Face Alignment via Cascaded Compositional Learning

Large-Pose Face Alignment via CNN-Based Dense 3D Model Fitting

CMS-RCNN: Contextual Multi-Scale Region-based CNN for Unconstrained Face Detection

Training Region-Based Object Detectors with Online Hard Example Mining

Joint Face Detection and Alignment Using Multitask Cascaded Convolutional Networks

Frontal to profile face verification in the wild

Deep Residual Learning for Image Recognition

SSD: Single Shot MultiBox Detector

WIDER FACE: A Face Detection Benchmark

From Facial Parts Responses to Face Detection: A Deep Learning Approach

A survey on face detection in the wild: Past, present and future

You Only Look Once: Unified, Real-Time Object Detection

Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks

Object Detection via a Multi-region and Semantic Segmentation-Aware CNN Model

Fast R-CNN

FaceNet: A unified embedding for face recognition and clustering

Joint Cascade Face Detection and Alignment

Edge Boxes: Locating Object Proposals from Edges

Aggregate channel features for multi-view face detection

300 Faces in-the-Wild Challenge: The First Facial Landmark Localization Challenge

Exemplar-Based Face Parsing

Face detection, pose estimation, and landmark localization in the wild

Annotated Facial Landmarks in the Wild: A large-scale, real-world database for facial landmark localization

Understanding the difficulty of training deep feedforward neural networks

Et al

Labeled Faces in the Wild: A Database forStudying Face Recognition in Unconstrained Environments

Face Detection: A Survey

Robust Real-Time Face Detection

TVM: An Automated End-to-End Optimizing Compiler for Deep Learning