Pattern Recognition and Machine Learning

ENSIMAG - 3 ( 5MM25371)

Grenoble Insititut National Polytechniques

Academic Year 2018 - 2019

James L. Crowley

Teaching Assistant:  Nachwa Aboubakr

Wednesdays 11h15 - 12h45 in  H203*

Verify Lecture Locations in ADE
Schedule of Classes

Class notes (pdf)

10 October 2018     Learning and Evaluation for Pattern Recognition - D208
                                Lab Project 1: Face Detection with  Color Histograms of Skin Pixels
                                (Oral Presentations   7 November,  Written reports due Wednesday 14 November)
                                UMASS FDDB: Face Detection Data Set and Benchmark
                                Background  Reading: [Jain and Learned-Miller 2010],  [Schwerdt-Crowley FG2000], (Nachwa Aboubakr)

17 October 2018  Detecting and Locating Faces with Color: Three Challenges for Project 1 - H202 (Nachwa Aboubakr)
                             Nachwa Aboubakr's ppt slides  on "Face Detection with Colors".

24 October 2018   Progamming Clinic:  Python, OpenCV, Face Image Data Bases  - H202  (Nachwa Aboubakr)

31 October -  pas de cours - Vacance de Toussaint

7 November 2018   Project 1 - Oral  Presentations by 2 teams -  H203

14 novembre 2018  No Class.  Project 1 written reports due by email.

21 novembre 2018   No Class

28 November 2018   Viola Jones Face Detector - Image Description with Haar like features, Boosted Learning, Cascade Classifiers.
                                   Lab Project 2: Viola Jones Faces Detector  (Evaluation Criteria)
                                  (Oral Presentations   12 December,  Written reports due Wednesday 19 December)
                                   Background Reading:  [Viola-Jones CVPR 2001]

05 December 2018 Clustering and Non-Supervised Learning with K-Means and EM
                                Jeff Bilmes, A Gentle Tutorial of the EM Algorithm

12 December 2018  Viola Jones Face Detector - Project Team Presentations

19 December 2018   Maximal-Margin Linear Classifiers: Perceptrons and Support Vector Machines
                                 Background Reading: Training Support Vector Machines: An Application to Face Detection

Alternative Lecture (not presented for lack of time).
19 December 2018   Eigen Faces - Face Detection and Recognition with Principal Components Analysis
                                 Background Reading: Turk and Pentaland, Face recognition using eigenfaces, CVPR 91

09 January 2019   Artificial Neural Networks:  Introduction, Multi-layer networks, Backpropagation.
                              Paper on Xavier GLORIOT's initialisation procedure
                              Background Reading: Face Dection with Neural Nets [Rowley-Kanade  1981
                              Lab Project 3: Face Detection with Artificial Neural Networks

16 January 2019   Convolutional Neural Networks, Pooling, Architectures
                              Background Reading: VGG, Simonyan and Zisserman, ICLR 2015
23 January 2019  Deconvolution, AutoEncoders, Generative Adversarial Networks
                            Background Reading:  Noh et al, Deconvolution for Semantic Segmentation. ICCV 2015
                            Background Reading:  Kingma et al, Deep Generative Models, NIPS 2014      

30 January 2019  Face Dection with Neural Networks  - Project Team Presentations
6 February 2018  Written  reports for Lab project 3  due by email

    Exam from  Jan 2017
    Exam from  Jan 2018

Face Detection Data Sets:

FDDB dataset: FDDB dataset contains the annotations for 5,171 faces in a set of 2,845 images.
WIDER FACE: A face detection benchmark dataset with 32,203 images and labels for 393,703 faces with a high degree of variability in scale, pose and occlusion.
MALF dataset: Face Detection in the Wild.  MALF consists of 5,250 images and 11,931 faces.
AFW dataset: Face Detetion in the Wild. AFW dataset is built using Flickr images. It has 205 images with 473 labeled faces.
For each face, annotations include a rectangular bounding box, 6 landmarks and the pose angles.

IJB-A dataset: US NIST IJB-A dataset for face detection and face recognition. IJB-A contains 24,327 images and 49,759 faces.
NB : Ces fichiers peuvent etre copié, reproduit et et inclus dans autre texte, sous condition d'inclure une citation :
These files can be copied and used in editing other text, with the condition of including a citation:
Copyright: Crowley, J. L. , "Class Notes - Pattern Recognition and Machine Learning",  ENSIMAG, Grenoble Institut Polytechnique, 

Crée par James L. Crowley. Last update 2 Dec 2018