Hi there, 1) I want doing a project for a school which requires some image processing. The main purpose, the PC with a camera, take a snapshot of a student. Then, it will convert the picture to jpg and do some image processing. Before doing image processing, it need to queried to School C picture database, and compare this student picture with the current picture. http://www.geocities.com/chuawenching/imagepr1.txt http://www.geocities.com/chuawenching/imagepr2.txt *copy the url to a new browser and then paste it. Geocities had some problem to access directly. Do you think this is the right way? I am worried, it will take some time to query the pictures from School C server. Do you think this is the right way to distribute the pictures from the central database to School C? Any help please? 2) Just wonder, what if the student image which is stored in the central database was taken 5 years ago, and when it took picture. Can it still manage to do the image processing? Just say, 5 years ago, he is a young boy face look. Now he had a matured look. Or maybe now he had a bald head. Any tips? Thanks. Regards, Chua Wen Ching [edited by - chuawenching on January 6, 2004 12:03:22 AM]

I am going to address the AI aspect of your problem, and not the networking issue..

I am also assuming that image processing means comparing to images of the same person..

Despite the endless ampunt of technologies that could help you out with comparing two facial images...it is very difficult for "computers" (meaning the programmers who develope on them) to detect enough facial features to compare against..anomalies that could go from; them going bald (indeed a good example on your behalf); how tan they are, differenct hair cut, whether they are wearing glasses or contacts now..etc, etc..

I had to search for in one of my A.I. Journals but I found the publication that will help you...

Publication:
Lecture Notes in Artificial Intelligence 1416: Tasks and Methods in Applied Artificial Intelligence - 11th International conference on Industrial and Engineering Applications of Artificial Intelligence
(don''t be scared at that long as title..lol)

Title: Automatic Storing and Retrieval of Large Collections of Images pg.629
Author: Agostino Poggi and Gianluca Golinelli

Also something that may help you with "image processing"
search for "gesture recognition eigen feature"

Let me know if this helps any

-Lucas

1) Thanks for the link. Hmm.. how do i find AI journals?

2) I will look into gesture recognition eigen feature. Eigen sounds familiar.

Does most face recognition engine use this concept now? Is it accurate and famous? I just want to confirm.

I heard eigen from someone before it is related to angles and sizes of an image. Is it accurate to compare images like this?

Thanks for the reply.

Regards,
Chua Wen Ching

I gave you things to help you search , Eigenspace representation in the simplest way i can think to explain it, would be representing the visual appearence of an object by not only its color/texture but also its object geometry.

To find journals
I use www.kluweronline.com but it may not be free to you depending on whether your highschool or university has an account with them..none the less you can still find journals and publication that are free somewhere. Just takes a while to find them.

You can keep in touch with me through email jxlxm@aol.com
but I would advise still using this board so that others can read & learn from this post aswell.

-Lucas

This forum is intended for discussing AI as related to games. We do occasionally permit off topic threads. For the time being I am happy to leave this thread open so long as the discussion maintains an AI theme. If you want to discuss aspects of networking or database performance, then please find a more appropriate forum/website to do this at.

Having said that, on the issue of face recognition...

Most recognition systems work on feature extraction. For faces though, this is not always what you might think of as a typical feature, e.g., a nose, the eyes, brow, etc. The sort of features that give good recognition results are things like the triangle formed by the eye centres and the point of the nose, the relative separation of the ears, height of the ear midline below the eyes, etc., etc. To implement this you would need to be able to segment an image of a face and then identify segments based on their position relative to other segments. Once you''ve done this, you can then extract the feature information based on the segment properties and finally perform a search based on the extracted feature description.

Once you want to start dealing with faces at angles though, it becomes a whole new ball game.

Good luck with your research.

Timkin

With all due respect, In defense for chuawenching, Genentic Algorithms and Neural nets are rarely used in Games, at best you see FSMs an a few DecTrees, because this topic is not used as frequently as others, and because alot the questions posted here are often hard to answer because more than half the people here don''t know much about A.I. and want to answer the question by saying "Why don''t you use a NN or GA"

I think this A.I. Board should be alot more relaxed concerning the questions that aren''t directly related to gaming,because there are alot of issues in the real world with A.I. that can inspire Game Developers. To say that this is an off topic thread, is insulting for chuawenching at best. We could easily debate that chuawenching implementation of his project has awesome and exciting uses in games, and may just give great ideas to Game Developers to this community.

Respectably,
-Lucas

Thanks. Anyway, i hope we don''t get ourselves into trouble.

I will look into the journal website you recommended. I will talked to my college to see whether i can get the journal you recommended.

Right now, i need to implement an agent that learnes. Well i do believe that eigenface uses somesort of neural network. Am i right?

The sort of features that give good recognition results are things like the triangle formed by the eye centres and the point of the nose, the relative separation of the ears, height of the ear midline below the eyes, etc., etc. To implement this you would need to be able to segment an image of a face and then identify segments based on their position relative to other segments.
--> Amazing. Never think until so detail. Cool!

Anyway again, i will do my own research too. Hehe!

Thanks Syntax and Timkin.

Regards,
Chua Wen Ching

quote:

---

Original post by Syntax
To say that this is an off topic thread, is insulting for chuawenching at best.

---

Lucas, please go back and re-read my post. I clearly stated that while the thread stayed on the topic of AI I was happy for it to remain open. My concern that it might not do this was based on a) the fact that it was not a game related thread (minor concern); and, b) that it was discussing a problem of image retreival and databasing, which is not directly related to AI (although such problems sometimes coincide with interesting AI problems).

Please don't jump down my throat simply because I am politely reminding the populace of the purpose of this forum. Yes, lately there have been quite a few threads that are off topic as far as game development is concerned... and that concerns me as many of our longer term, regular contributors have stopped contributing. I have been happy to let these threads persist because I want to keep this community going... I want to see people participating every day... and hopefully, with a little gentle encouragement, we can spend more time discussing games and interesting game AI, rather than debate the merits of ANNs, GAs or other buzz-word technologies.

If it appeared that I was jumping down chuawenching's throat, then you have my apologies. That was not the intention. I hope that in re-reading my post you come to understand that and recognise that I am not here to regiment and restrict this forum, but rather to help guide it and maintain it as an interesting and exciting place that we can all visit each day.

Regards,

Timkin

[edited by - Timkin on January 8, 2004 8:52:39 PM]

Don''t worry about it''s all cool...thanx though for allowing this thread to persist.

-lucas