ITCS 6157/8157 Visual Database Fall 2015 http://www.cs.uncc.edu/~jfan/itcs6157.html
Dec 30, 2015
ITCS 6157/8157
Visual Database
Fall 2015
http://www.cs.uncc.edu/~jfan/itcs6157.html
Overview
Class hour 9:30AM - 12:15PM, Thursday Office hour Thursday 1:30 - 5:00PM Classroom Woodward Hall 130 Instructor - Dr. Jianping Fan
email - [email protected] Office – Woodward 205D Webpage http://www.cs.uncc.edu/~jfan
Textbook: we will use the slices and papers on the course web page
Why we should have this course?
Internet is changing the world Multimedia (especially visual data) is
dominating the content of Internet Easy access of multimedia content
through Internet could be the future of IT
This class will provide training on multimedia content analysis and search!
Why we should have this course?
Good job market: Google, Yahoo!....
Have fun: solving real problem
Not so “hard” to learn (??)
Next generation search engines
How can I access multimedia in database over networks?
Networks
Course objectives
Multimedia ServerUser
Google, Yahoo! & MSN IE
1. How to format multimedia queries? 2. How to represent multimedia content?3. How to index large-scale multimedia?
4. How to search multimedia in database ? 5. How to transmit query results over IP ?6. How to control user’s access ?
To answer above question, we need to address:
Course objectives
Everyone has smartphone now
How to build multimedia search engines?
How to build text indexing?Text document
Natural language processing
Inverse File indexing
Text database
Multimedia data
multimedia analysis
Hash Indexing or othersMultimedia database & query
Yahoo, Google
Can we do multimedia retrieval like Google for text search?
Keywords
Multimedia ``keywords”
Simple extension
Required Techniques How to build multimedia search engines?
Computer Vision Technologies for Multimedia Content Analysis
Machine Learning Tools for Understanding Multimedia Semantics
Database Techniques for Large-Scale Multimedia Indexing
Human-Computer Interaction for query formulation, display & exploration
Components from Database System
a. Data Representation Schema
b. Database Indexing
c. Database Storage
d. Query Management
e. Big Data Analytics
Components from Computer Vision
a. Image & Video Analysis & Feature Extraction
b. Object Detection & Scene Understanding
c. Classifier Training for object and concept detection
d. Scene Configuration and Structure
Components from Machine Learning
a. GMM & Bayesian Network
b. Support Vector Machine (SVM)
c. Graphical Models & Structure Learning
d. Statistical Inference
e. Deep Learning & Big Data Analytics
Database Management System: ITCS6160 or ITCS3160
Computer Vision Machine Learning Programming Skills
Willing to work hard
If you do not have these background, you should
Course Topics
Data Clustering Tools Machine Learning Techniques Multimedia Analysis Technologies Database Indexing Structures Deep Learning & Big Data Analytics Human-Computer Interaction Tools Taking-Home Self-Study Materials Open Discussion & Student Presentation
Grading
Composition Project 25% Midterm 35% Final 40%
Scale >93% = A 75-93% = B 55-74% = C <55% or cheating = F
If you miss 3 classes (three weeks) or more, you are not allowed to take tests (mid-term and final)!
Class Policy You have to attend the class & come to
classroom on time (no later than 9:35am)
You should be ready to learn from the class: project implementation could be critical
You should respect your classmates: come to learn from their presentation!
Classroom Policy
No food!!! Drink could be allowed & Cell Phone should be turned off.
Small talk is not allowed, but you are welcome to ask questions!
Walking inside classroom is not allowed within presentation time!
Course Projects
Project implementation project: you need to set up a team or individual to implement one small system for multimedia content analysis or understanding.
Paper presentation project: you need to pick one topic to present in the class. MS students are not encouraged to take this kind of project!
More information http://www.cs.uncc.edu/~jfan
We will offer two kinds of projects:
Implementation Project
Develop image/video analysis system using Visual C++ and Java. Each group consists 3-4 students 3-4 hours workload each week is expected Java or C++ assumed
Talk to your professor to decide which algorithm you may implement for your project, discuss progress with your professor if necessary
Demonstrate your implementation to your professor & get feedback
Paper Presentation Project Present one research topic: you need to
talk to your professor to get relevant research papers, prepare presentation slides & present in the class. Well-understanding of the topic Good presentation in the class Be able to answer questions from classmates
& professor Topic selection: depending on available
topics and professor assignment.
If you are PhD student
Course Projects
Good grade even you may perform well in final and mid-term tests Practical implementation means more than
paper work
Good recommendation letter for job hunting: professor can only memorize students with good performance!
Research position opportunities
If you do wonderful job on course project, you may expect:
Midterm & Final
closed books and notes One page notes is permitted
Cumulative No makeup Bonus is expected Key components for your final
gradeIf you miss 3 classes or more, you are not allowed to take tests (mid-term and final)!
Suggestions from Instructor
Do your best in the class Show your problems to the
instructor when you cannot make it Show the evidence to us if you
think you are right. Open discussion is welcome, but no
small talk
10-hours Golden Rules 3 hours before class: go through the topics,
presentation slides and seek some relevant online documents, …; ready to ask questions in class
3 hours in class: listen to domain experts and try to ask questions
4 hours after class: review what you have learnt from the class, do your project and assignments…
Who cares?
Who cares?
Google Search Engine
Google Search Engine
Who cares?
The way to join them
Good grade from class
More training on programming skills, especially for multimedia analysis, indexing and retrieval
Get recommendation from professor
Recommendation
Good grade is very important, but it is not everything!
Learning something and solving one problem you like may be more important!
Learning from someone who may make you better! Especially your classmates
What areas we will touch?
Computer Vision Database Information Retrieval Machine Learning & AI Visualization Networks Statistics & Security
What you may expect
Start-up Companies
Many wonderful companies & start-ups come from course projects!
You could be the next one!
What you may expect
Start-up Companies
Product search engine for amezon.com, taobao.com
Using your smartphones to take pictures, then we will find the cheapest ones for you!
What you may expect
Start-up Companies
Image Search Engine
What you may expect
Start-up Companies
Google Glass App: Google glass may change world like i-phone
What you may expect
Start-up Companies
Digital Camera App: Sony may sale digital cameras with your media organization & searchsoftware.
What you may expect
Start-up Companies
Personal Computer App: IBM Dell may sale PCs with your media organization & search software.
What you may expect
Start-up Companies
Automatic-Driving Car App: BMW Tesla may sale cars with your object recognition & navigation systems.
What you may expect
Start-up Companies
Plant Species Identification: Your Kids will be proud of you because you Know every plant species on the world
What you may expect
Start-up Companies
Plant Species Identification: Your Kids will be proud of you because you Know every plant species on the world
What you may expect
Start-up Companies
Construction Safety Alarm: educators, government & insurance companies may care
What you may expect
Start-up Companies
Multimedia Search Engine:Google will definitely care!
What I or UNCC may expect
Do not forget to come back UNCC & support our research!
Do your best & have fun!
Good students should be able to push your professor to think and work harder not easier!