Extensive Facial Landmark Localization with Coarse-to-fine Convolutional Neural Network Erjin Zhou Haoqiang Fan Zhimin Cao Yuning Jiang Qi Yin Megvii Inc.,

Slides:

Advertisements

Similar presentations

Foundations of Technical Analysis: Computational Algorithms, Statistical Inference, and Empirical Implementation Written by Andrew W.Lo, Harry Mamaysky,

Advertisements

Part II – TIME SERIES ANALYSIS C3 Exponential Smoothing Methods © Angel A. Juan & Carles Serrat - UPC 2007/2008.

Support Vector Machines

Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

|| Dmitry Laptev, Joachim M. Buhmann Machine Learning Lab, ETH Zurich 05/09/14Dmitry Laptev1 Convolutional Decision Trees.

Message Splitting Against the Partial Adversary Andrei Serjantov The Free Haven Project (UK) Steven J Murdoch University of Cambridge Computer Laboratory.

Spike Sorting Goal: Extract neural spike trains from MEA electrode data Method 1: Convolution of template spikes Method 2: Sort by spikes features.

High-level Component Filtering for Robust Scene Text Detection

Automatic Pose Estimation of 3D Facial Models Yi Sun and Lijun Yin Department of Computer Science State University of New York at Binghamton Binghamton,

Computer Vision Laboratory 1 B-Spline Channels & Channel Smoothing Michael Felsberg Computer Vision Laboratory Linköping University SWEDEN.

I.1 ii.2 iii.3 iv.4 1+1=. i.1 ii.2 iii.3 iv.4 1+1=

Integration Testing Presented By Nesreen Ahmad. *Main Points:-  Definition Of Integration Testing.  Procedure Of Integration Testing.  Integration.

I.1 ii.2 iii.3 iv.4 1+1=. i.1 ii.2 iii.3 iv.4 1+1=

Smart Traveller with Visual Translator for OCR and Face Recognition LYU0203 FYP.

Chapter Three Research Design.

General Mining Issues a.j.m.m. (ton) weijters Overfitting Noise and Overfitting Quality of mined models (some figures are based on the ML-introduction.

Facial Recognition CSE 391 Kris Lord.

Eigenfaces for Recognition Student: Yikun Jiang Professor: Brendan Morris.

DOOR ENTRY SYSTEM Alina Dinca László Papp Adrian Ulges Csaba Domokos Cercel Constantin Team:

Rotation Invariant Neural-Network Based Face Detection

GAZE ESTIMATION CMPE Motivation  User - computer interaction.

LeCun, Bengio, And Hinton doi: /nature14539

8-1 Chapter 8: Image Restoration Image enhancement: Overlook degradation processes, deal with images intuitively Image restoration: Known degradation processes;

Motion Estimation I What affects the induced image motion?

SAFE KERNEL EXTENSIONS WITHOUT RUN-TIME CHECKING George C. Necula Peter Lee Carnegie Mellon U.

Animation From Observation: Motion Editing Dan Kong CMPS 260 Final Project.

3D head pose estimation from multiple distant views X. Zabulis, T. Sarmis, A. A. Argyros Institute of Computer Science, Foundation for Research and Technology.

基督再來（一）. 經文： 1 你們心裡不要憂愁；你們信神，也當信我。 2 在我父的家裡有許多住處；若是沒有，我就早已告訴你們了。我去原是為你們預備地去。 3 我若去為你們預備了地方，就必再來接你們到我那裡去，我在那裡，叫你們也在那裡， ] ( 約 14 ： 1-3)

Face Detection 蔡宇軒.

Facial Smile Detection Based on Deep Learning Features Authors: Kaihao Zhang, Yongzhen Huang, Hong Wu and Liang Wang Center for Research on Intelligent.

Face Recognition based on 2D-PCA and CNN

FINGERTEC FACE ID FACE RECOGNITION Technology Overview.

Facial Detection via Convolutional Neural Network Nathan Schneider.

When deep learning meets object detection: Introduction to two technologies: SSD and YOLO Wenchi Ma.

Pining for Data II: The Empirical Results Strike Back

CNN-RNN: A Uniﬁed Framework for Multi-label Image Classiﬁcation

Network Scale-up method

Implementing Boosting and Convolutional Neural Networks For Particle Identification (PID) Khalid Teli .

Ajita Rattani and Reza Derakhshani,

Motivation Computers are good at some things… Calculating 

Dynamic Routing Using Inter Capsule Routing Protocol Between Capsules

A Convolutional Neural Network Cascade For Face Detection

ФОНД ЗА РАЗВОЈ РЕПУБЛИКЕ СРБИЈЕ

استخراج فلزات 1 آماده‌سازی بار. استخراج فلزات 1 آماده‌سازی بار.

NeC4.5 —— Neural Ensemble Based C4.5

Object Detection + Deep Learning

Слайд-дәріс Қарағанды мемлекеттік техникалық университеті

Hairong Qi, Gonzalez Family Professor

.. -"""--..J '. / /I/I =---=-- -, _ --, _ = :;:.

8-3 RRAM Based Convolutional Neural Networks for High Accuracy Pattern Recognition and Online Learning Tasks Z. Dong, Z. Zhou, Z.F. Li, C. Liu, Y.N. Jiang,

II //II // \ Others Q.

'III \-\- I ', I ,, - -

I1I1 a 1·1,.,.,,I.,,I · I 1··n I J,-·

Networks and the Internet

Forward and Backward Max Pooling

Emna Krichene 1, Youssef Masmoudi 1, Adel M

The 9 Deep Learning Papers You Need To Know About (Understanding CNNs Part 3) Fait un tour historique du domaine: quels articles/travaux ont été marquants.

Heterogeneous convolutional neural networks for visual recognition

Multi-Information Based GCPs Selection Method

DATABASES AS MINDTOOLS

Counting Statistics and Error Prediction

Counting Statistics and Error Prediction

. '. '. I;.,, - - "!' - -·-·,Ii '.....,,......, -,

Department of Electrical Engineering

Types of Errors And Error Analysis.

Weekly Learning Alex Omar Ruiz Irene.

REU Week 1 Jared Rhizor.

Presentation transcript:

Extensive Facial Landmark Localization with Coarse-to-fine Convolutional Neural Network Erjin Zhou Haoqiang Fan Zhimin Cao Yuning Jiang Qi Yin Megvii Inc., Beijing

Problem 2-Eyes Detection

Problem 2-Eyes Detection5-Corners Detection

Problem 23-Points Detection

Problem 68-Points Detection More and more landmarks are required!

Idea I Single Predictor Disadvantage: the difficulty of localizing each point is quite different, and it is hard to optimize all points by a single model.

Idea II Predictor 1 Predictor 2 Predictor 3 Predictor 4 …… Predictor 5 Disadvantages: no geometric constrains; heavy computational burden.

Idea III Predictor 1 Predictor 2 Predictor 3 Predictor 4 … Advantage: component contexts are considered.

Observation However, do we really need the mouth to locate the eye corners?

Our Intuition Component estimator Predictor 1 Predictor 2 Predictor 3 Predictor 4 …

Framework

Experiments

Level 2Level 3Level 4 Average error on each level of CNN.

Results on 300-W

About us

Thanks!