+ All Categories
Home > Documents > Keep Smiling! GRIM GRINS. The Project’s member György Hingyi – programmer & manager1 Péter...

Keep Smiling! GRIM GRINS. The Project’s member György Hingyi – programmer & manager1 Péter...

Date post: 30-Dec-2015
Category:
Upload: helena-george
View: 215 times
Download: 0 times
Share this document with a friend
Popular Tags:
12
Keep Smiling! GRIM GRINS
Transcript

Keep Smiling!GRIM GRINS

The Project’s member

György Hingyi – programmer & manager1

Péter Szabó – programmer & manager2

Sinan Oz – scientist

Krisztina Maróti – do „paper” work

The problem

• Smiling faces.• Input: a set of photos of the same person

with different face expressions and the information that some of them are smiling faces 

• Task: to write a program (e.g., neural network) recognizing the smiling faces of the same person

• Output: smiling or not, and the statistics of the implemented method. 

• Difficulty: hard

Method Selection

•K-nearest neighbors

•Neural networks

•Ensembles of neural network classifiers

•Set of experts

•Support Vector Machine

•Other methods – Other New brainchild

•Etc…

Other’s works

• Real-Time Emotion Recognition using Biologically Inspired Models Keith Anderson, Peter W. McOwan – Using SVM

• Recognizing Emotion From Facial Expressions: Psychological and Neurological Mechanisms Ralph Adolphs University of Iowa College of Medicine – Using Others

• Recognizing Emotion in Speech, F. Dellaert, Proceedings of the ICSLP '96, October, 1996. – Try a lots of method and use combined one.

• Emotion Recognition and Its Application to Computer Agents with Spontaneous Interactive Capabilities - Ryohei Nakatsu, Joy Nicholson and Naoko Tosa – Using Neural network

„Our” problem

• Smiling faces.• Input: a set of photos of the not same person

with different face expressions and the information that some of them are smiling faces 

• Task: to write a program (with SVM) recognizing the smiling faces (on video) of the not same person

• Output: smiling or not, and the statistics of the implemented method. 

The SVM

• Support Vector Learning

Our Solution

1. step We need lots of pictures:

2. Step Processing the pictures

Our Solution

17x17

22x22

28x28

36x36

Our Solution

4. Input : video file

detect face(s)

detect emotion

„Future” Work – TO DO

What we wanted to do, but we have not enough time…

Tesekkurler

Thanks for your attention

Köszönöm a figyelmet

Danke schön

Vã multumesc foarte mult!


Recommended