fanap internship

Post on 16-Apr-2017

170 views 1 download

Transcript of fanap internship

FANAP SUMMER INTERNSHIP1394

Hanzaleh Akbari Nodehi – EE 91

Ghazal Helmzadeh - CS 92

Alireza Amirshahi – EE 91

Pardis Pashakhanloo - CE 91

Pourya Mandi Sanam – EE 91

Amir Hossein Nazem – EE 91

OUR TEAM

INTERNSHIP: AN OPPORTUNITYOR just another wasted labor?

◦Getting hands dirty

◦Boosting skills

◦Broadening horizons

◦Dude, let me tell you something…real world is much bigger than university!

◦Realistic decisions based on real interests

DO YOU REMEMBER THE FIRST DAY YOU ENTERED THE UNIVERSITY?!

We started learning…

Open-Source Computer Vision Library

No more reinventing the wheel

OPENCV

Opencvmodules

core

imgproc

objdetect

video

ml

highgui

Opencv seems better:

• Money

• Speed & Performance

• Cross-platform

• Ease of Use

FACE DETECTION AND FEATURE EXTRACTION

FACE DETECTION

First approach: skin color

Machine learning… better job

PAUL VIOLA & MICHAEL JONES

WE NEED FEATURES…

MAIN IDEAPositive

And Negative Images

Extract feature

Integral image

Features with

minimum error

Weighted sum of

features

Cascade

OBJECT DETECTION

Ball Detection with opencvObject Detection Using Color

Edge Detection

Tracking Object

Let’s see …

Machine learning

In 1959, Arthur Samuel defined machine learning as a "Fieldof study that gives computers the ability to learn withoutbeing explicitly programmed".

Arthur Lee Samuel was an American pioneer in the field of computer gaming, artificialintelligence, and machine learning.Born: December 5, 1901, Emporia, Kansas, United StatesDied: July 29, 1990, Stanford, California, United StatesEducation: College of Emporia (1923), Massachusetts Institute of Technology

WHAT IS MACHINE LEARNING ?!!

without being explicitly programmed !!!

If (this expression is true) {code block

} Else {another code block

}

Let me explain it with examples …

How do we learn ?!!

Time …

???

. . .

That was four.

• Machine Learning on the other hand is writing aprogram that learns from past experience.

It’s Applications …You've probably use a learning algorithm. dozens of times a day without knowing it.

Web Engines Photo Tagging Applications Spam Filter…

Web Engines Photo Tagging Applications Spam Filter…

EVERY TIME YOU USE A WEB SEARCH ENGINES LIKE GOOGLE OR BING TOSEARCH THE INTERNET, ONE OF THE REASONS THAT WORK SO WELL IS BECAUSEA LEARNING ALGORITHM, ONE IMPLEMENTED BY GOOGLE OR MICROSOFT, HASLEARNED HOW TO RANK WEB PAGES.

It’s Applications …You've probably use a learning algorithm. dozens of times a day without knowing it.

Web Engines Photo Tagging Applications Spam Filter…

Web Engines Photo Tagging Applications Spam Filter…

EVERY TIME YOU USE FACEBOOK OR APPLE'S PHOTO TAGGING APPLICATIONAND IT RECOGNIZES YOUR FRIEND'S PHOTOS, THAT'S ALSO MACHINELEARNING.

It’s Applications …You've probably use a learning algorithm. dozens of times a day without knowing it.

Web Engines Photo Tagging Applications Spam Filter…

Web Engines Photo Tagging Applications Spam Filter…

EVERY TIME YOU READ YOUR EMAILS AND YOUR SPAM FILTER SAVES YOU FROMHAVING TO WADE THROUGH TONS OF SPAM EMAIL, THAT'S ALSO A LEARNINGALGORITHM.

Future …

For big companies like facebook, google, yahoo and …, one of the reasons they're excited is theAI dream of someday building machines as intelligent as you or me.

Reading a captcha

How to read a

C A P T C H A

What is captcha?An acronym for "Completely Automated Public Turing test to tell Computers and Humans Apart”

A CAPTCHA is a type of challenge-response test used in computing to determine whether or not the user is human.

What we have done?Different approaches to read characters automatically

Optical Character Recognition

Decision Tree

Neural Network

Template Matching

Etc.

What we have done? (Cont’d.)Developing a NN based program to automatically decipher a Captcha

Advantages of using Neural Networks (our approach) for character recognition

Robust against noise and change in font, size, etc.

Easily portable to other fonts, characters, and even languages.

NowLet’s

down a server!

Motion detection

Motion Detection – Simple Approach

Using a threshold◦Problem: most cameras produce noisy images

◦Solution: Erosion filter

Motion Detection – Common Approach

Compare frames◦Good idea with compression in mind◦Disadvantage: slow movement

Motion Detection – Better Approach

Change the way you compare◦Independent from motion speed◦Disadvantage: dealing with gone objects

Motion detection - Most Efficient Approaches

Based on building the background of the scene

THANKS!

the end