[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference abbott::neural-networks

Title:The New Neural Networks Conference
Moderator:LGP30::FLEISCHER
Created:Fri May 25 1990
Last Modified:Sun May 04 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:82
Total number of notes:276

80.0. "Does anybody still read this conference?" by SUBSYS::POIRIER () Fri Mar 28 1997 15:35

Hello,

Has everybody lost interest in this subject?  I am interested in
neural nets for use with image recognition.  I have been working
on taking bitmap images from a camera and feeding them into
a network for real time image recognition.
T.RTitleUserPersonal
Name		DateLines
80.1still interested ....MXOC00::JGUTIERREZThu Apr 03 1997 13:3610
    Hi,
    
    I am still interested in this stuff. Anybody else? it looks like
    everybody is on other non-related-ai task! maybe so.
    
    By the way, have you taken a look to the Digital Cambridge Reasearch
    group. It sounds like they're working on image recognition.
    
    Jorge.
80.2Suprise, suprise.. They are still out there.SUBSYS::POIRIERThu Apr 10 1997 00:5021
I have not heard of anything the Cambridge Research Group is doing..
I would be curious to know what types of results they are getting though.
They must be using special hardware?  Connectionist machines?   
What I have been finding is that there are several drawbacks to using
neural nets.  
One, I can't seem to get the network to learn what I want it to.  For
instance I can feed it large amounts of training examples and it will
learn to respond positively to that image but as soon as it sees something
new, results are unpredictable.  There are so many variables involved as well,
such as sizes of layers..
Two, the amount of time to do even the SMALLEST images takes forever.
I am using a P133 with 50MB of RAM and trying to train the network
on an image of 160x120 (19200 pixels in input layer) maxes out the CPU
at 100% and can go on for days.

Anyway, I will try and find some info on the Cambridge group, if you know of
a pointer that would be cool..

Thanks for the response.

/Jamey
80.3Cambridge Research Lab pointer...MXOC00::JGUTIERREZTue Apr 15 1997 15:1511
    Hi Jamey,
    
    A good point to start taking a look follows:
    
    http://www.crl.dec.com/home.html
    
    good luck!
    
    Jorge.
    PS. maybe you will let us know what you find out....