Webcam Face Recognition

[moreis62]

We should start by defining all the product functionality, and specifications. But Don't forget we need to developed plug-ins to be run with the thinking process which implements all aspects of Hal's behavior from low level feedback learning control such as what we'er doing here on this forum and at the iCat site.  Additionally, Mr.Medeksza should devised and implemented procedures for quantifying coverage performance for fine-tuning of control parameters and comparison with competitor's products like he did with Haptek, just for in emergency. just in case some thing goes wrong with Hal and we can't control him. Now from what i've tested at (DARPA), (x10) i'ev don in Ultra Hal, but it would be wize and really neet to Afillyeat facial recognition in to a webcam by using Either
 
The ERSP 3.1 Software Development Kit by Evolution Robotics,
http://www.evolution.com/products/ersp/
 or
http://www.hitech-projects.com/icat/index.php
The iCat platform that is made here at Philips Research and tested to work with facial recognition.

We can all customize any of Ultra Hal's robotic solutions to fit your product's Needs Mr.Medeksza.

Videos on iCat: My take some time to Download
but here it is finely.

http://www.hitech-projects.com/icat/download/iCat-TV-Assistent-and-Personality.wmv

http://www.hitech-projects.com/icat/download/iCat-Game-Buddy-Katy.wmv


http://www.hitech-projects.com/icat/download/iCat-Game-Buddy-Compare.wmv

[freddy888]

quote:Originally posted by GamerThom
Freddy, You must not have tried talking to Hal via headset
mic with it running full screen. Try it, it'll give you
the shivers. []

Nope, a headset and mic is on my list of things to buy still but I have a car that needs money spent on it this month, so it will have to wait a while yet.  It will happen though [8D]

On all this integration I am still in the dark on how to go about doing a lot of it, I did have an idea for some kind of 'hub' program that could relay/distribute info from any number of other programs to and from Hal - but I don't know if that is necessary or not??  You'd have the hub program basically as an information gatherer and sorter that Hal would refer to when required...

It's a similar idea to that used in a music/mixing program that can control a variety of differing inputs - midi, wav, syth-boxes and signal generators and mix them all into a final result, also being able to do it all in realtime too.

[Art]

I think that the webcam software could be integrated with Hal via a plugin script, perhaps using the active-x.

Finding one that works as hoped in another issue.

Also, IF we are able to get video recognition working with Hal, would we need a routine to allow Hal to observe facial positions or concentrate on chatting? Would Hal keep trying to identify the user while it's chatting with us? How many times or is once enough? What if users were switched? Would the constant barrage of video imagery overload Hal (or our computer's memory)?

Acquiring the software and interfacing it to work with Hal is only the tip of the iceberg, so-to-speak.

[freddy888]

Yep good points, on the overloading Hal, that is partly why I had the idea of the hub program.  Essentially leaving each program to its own devices, but having this kind of mediator relationship going on between the programs.  That way I figured Hal wouldnt bloat out too much and it might be easier to involve further programs/systems that way.  Thats probably getting too much for a normal PC though I think, but distributed computing might be part of the answer...

That Jeramiah works over several PC's it seems, but I am not sure....hmm to blow up my pc or not to blow up my pc that is the question..

[Art]

Actually, Hal's routine for visual identification would most likely be placed at the beginning of the program, where the user usually addresses Hal with, "Hi", "Hey there", etc. then stop.

Imagine the possibility of constant recognition however, similar to iCat wherein it constantly sees things, receives info and responds with emotional displays.

2 sides to every coin....

[GamerThom]

What if the facial recognition routine were to be run as temporary
memory After initial identification of user, with the temporary
image acquisitions to memory being deleted by a subroutine, let's
say every 3 minutes. That would allow the aquisition of new image
information with out overloading Hals program or the host system.

[Art]

Not a bad approach, GT.

Yesterday I came across a site that used a webcam
to view a user's head gestures to control the cursor
as well as left / right click and scroll!
Pretty neat technology out there.

If only we could find some decent vision software
for Hal.

[GrantNZ]

Crikey. You wouldn't want to sneeze.

One annoying problem with some alternate computer inputs is that they can't filter out unintentional inputs. Voice recognition software sometimes misinterprets something as a command, and suddenly it's done something and you don't always even know what. I have a PocketPC, which has a touch screen, and brushing a crumb off the screen can cause hell to break loose.

Video control of the cursor.... Imagine if you had to swat a mosquito or something. I can just imagine a mish-mash of inputs, followed by an (intelligent) Hal responding "is there an error with your brain?" []

It'd be curser controlling cursor....

Anyway, this isn't contributing much, excuse me []

[freddy888]

No that's actaully a very good point, you really don't want Hal doing things you don't want especially if you get to the point where he is even more integrated.  That gives me a further idea for my hub thing - the hub program should be able to filter out things that Hal is not supposed to process.

[GamerThom]

Freddy, You referring to scripting what input Hal will
recognize, only acting/reacting to specific hand gestures
and facial expressions. Similar to operant conditioning in
animals and people, but achieved through program scripting.
Wouldn't that put unwanted limits on the Bot, if what is
trying to be achieved in the first place is a Bot with the
capabilities of exploring the world and interacting with us?

[Art]

It would be kind of neat if Hal via the webcam would see the user smile or laugh and be able to know the difference and respond accordingly. Such reactions would really bring new meaning to artificial intelligent beings.

<I know you were talking about disconnecting me...I could see your lips moving> ...to paraphrase Hal from 2001.[]

[freddy888]

GamerThom, I was just thinking more along the lines of preventing crashes and things. Pre-processing the input to check that nothing crashes the system, not so much to limit whatever Hal could do.

*stumbles in the dark again oh there's the oil-lamp*

>LIGHT LAMP

>The lamp lights and you see passageways east, north and west.  From the North comes the sound of engineers figuring out artificial brains, from the west are the sounds of partying fullbodies.  From the east there is the sound of a small stream.

[]

[aladyblond]

oh ma gawd! is that the ramblings of an ancient zork program? i loved zork the old typed one. i played that so many times...~~alady.

[GamerThom]

[^] Aah, OK, got it.

[?]What's coming from the South?[]

[aladyblond]

from the south are the sounds of a new frontier the virtual world,with virtual time where it is the same time everywhere and a.i.s are working fervently on creating the first human.....(got carried away,[:p] sowwy.).~~alady