[Memo] 12/1

Comparision between Algorithms:

• edge detection，skin color及motion等方法都可以很快找到人臉範圍，但較容易受到外界環境干擾。
• PCA，LDA，FA、SVM及neural network等方法雖然有不錯的偵測率，但需花費很長的時間。
• A Face-Finding Search Engine: http://www.technologyreview.com/communications/21384/?a=f for low-resolution images."Super-resolution will give you an interpolated image that looks better," says Hennings-Yeomans, "but it will have distortions like noise or artificial [features]."
• Face recognition: http://image.cse.nsysu.edu.tw/research/Delaunay/delaunay.htm Advantage:unique, local re-meshing, arrangement invariance and can detect edges. 可利用三角形結構作為類神經網路的輸入。
  •人臉資料庫由麻省理工學院媒體實驗室(MIT Media Laboratory)取得http://vismod.www.media.mit.edu/vismod/demos/photobook (144位不同人正面影像，每個人有二到四張不同之臉部正面影像，共有308張影像)

---------------------------------------------------------------------

人臉識別發展: http://140.113.87.114/cvrc/edm/vol_7/tech1.htm
人臉辨識相關網誌: http://www.im.ntu.edu.tw/~d93001/blog/index.php?d=06&m=11&y=08&category=4
人臉辨識系統(流程與演算法): http://140.128.102.71/research/mn/02/index.htm

openCV學習筆記與心得(含程式範例): http://funp.com/t136721#p=136721
openCV自學手冊: http://vision.csie.ncku.edu.tw/~euleramon/openCV_DIY.pdf

橢圓人臉偵測+膚色偵測+轉黑白+轉輪廓(含程式範例): http://delphi.ktop.com.tw/board.php?cid=31&fid=79&tid=73445

[Meeting] 11.27

Topic: mypikaray, a social networking website with face recognition and retrieval

Objective: Search for people by photos

Methods:
1. CBIR
- Global color space
- Garbor
- Search for what feature is suitable for face identification
- Ask Winston
2. OpenCV
3. php+mySQL+Apache
4. Website design
5. jQuery

Resources:
1. AppServ:
http://sourceforge.net/project/downloading.php?groupname=appserv&filename=appserv-win32-2.5.10.exe&use_mirror=jaist
2. jQuery:
http://www.jquery.com.tw/
3.OpenCV:
http://www.opencv.org.cn/index.php?title=%E9%A6%96%E9%A1%B5&variant=zh-tw
4. Face recognition:
http://image.cse.nsysu.edu.tw/research/Delaunay/delaunay.htm

Task:
1. Survey the CBIR algorithm on face recognition
2. Study and test OpenCV

Next meeting:
12/2(二)

[Reference Article] A Face-Finding Search Engine

http://www.technologyreview.com/communications/21384/?a=f

Open source album system

Here are some websites about open source album system.
And it might be useful when building our site.

http://www.phpalbum.net/

http://www.plogger.org/

http://gallery.menalto.com/

http://www.econsultant.com/i-want-open-source-software/index.html

http://www.techsupportalert.com/content/open-source-album-art-finder.htm

Project idea

Current idea:
We'll build a simple social networking website that includes the album system. The key feature of the album system is automatic classification. If users want to upload lots of photo to the album system, they just click the upload button, and the website will classify all these photo automatically based on CBIR technique.

Besides, another feature of this social networking website is that users can search other user's profile just by photo.

However, the project idea have not proposed completely.
And we'll keep discussing and refining our thought.

2008 Fall MMAI

This blog is for final project of MMAI in Fall, 2008.
Our team is mypikaray.
And the team members are Chien-Ray Kung, Yi-Chun Chen, and Heng-Yu Chi.

Course website: http://www.csie.ntu.edu.tw/~winston/courses/mm.ana.idx/
Lecturer: Winston Hsu