Post by thread:[OT] TRANSPUTERS

Exact match. Not showing close matches.
PICList Thread
'[OT] TRANSPUTERS'
2000\04\27@131059 by WF

Here can i find low cost TRANSPUTERS? My friend developted a C++ program that recognizes HUMAN FACES...He have tested in many angles the FACE and worked very fine...ERRORS: 0%. He uses NEURAL NETWORKS... The system captures an image of my camera, my face for example, is done a Trainning on the NEURAL NETWORKS and ready, it's disponible for recogzine my Face in all angles... I suggested to put this Software in a EMBEDDED SYSTEM, the problem is that, the C++ Software take +-5 minutes to recognize the image... :( in a 266Mhz computer...) Well, i suggested to use a EMBEDDED TRANPUTERS for Parallel Mathematical Computing...We intend to use a Hardware that take +-2 second for recognize... What do you, PICLIST USERS, think about? Miguel Wisintainer Master Computation

2000\04\27@143638 by M. Adam Davis

Transputers are good for parallel processing, perhaps they would be a good target for your app. However, they are not commonly used, and could be about as expensive to devlop for as it would cost to simply get a 1GHz 64-bit processor (alpha, or other). Perhaps you could even use a custom OS on a dual-processor pentium system. Personally I wouldn't look at PIC processor to get the job done ;-) -Adam WF wrote: {Quote hidden}> > Here can i find low cost TRANSPUTERS? > > My friend developted a C++ program that recognizes HUMAN FACES...He have > tested in many angles the FACE and worked very fine...ERRORS: 0%. He uses > NEURAL NETWORKS... > The system captures an image of my camera, my face for example, is done a > Trainning on the NEURAL NETWORKS and ready, it's disponible for recogzine my > Face in all angles... > I suggested to put this Software in a EMBEDDED SYSTEM, the problem is that, > the C++ Software take +-5 minutes to recognize the image... :( in a 266Mhz > computer...) > Well, i suggested to use a EMBEDDED TRANPUTERS for Parallel Mathematical > Computing...We intend to use a Hardware that take +-2 second for > recognize... > > What do you, PICLIST USERS, think about? > > Miguel Wisintainer > Master Computation

2000\04\27@143853 by WF

Yes, yo're correct! Miguel ----- Original Message ----- From: M. Adam Davis <spam_OUTadavisTakeThisOuTUBASICS.COM> To: <.....PICLISTKILLspam@spam@MITVMA.MIT.EDU> Sent: Thursday, April 27, 2000 3:35 PM Subject: Re: [OT] TRANSPUTERS > Transputers are good for parallel processing, perhaps they would be a good > target for your app. However, they are not commonly used, and could be about as > expensive to devlop for as it would cost to simply get a 1GHz 64-bit processor > (alpha, or other). Perhaps you could even use a custom OS on a dual-processor {Quote hidden}> pentium system. > > Personally I wouldn't look at PIC processor to get the job done ;-) > > -Adam > > WF wrote: > > > > Here can i find low cost TRANSPUTERS? > > > > My friend developted a C++ program that recognizes HUMAN FACES...He have > > tested in many angles the FACE and worked very fine...ERRORS: 0%. He uses > > NEURAL NETWORKS... > > The system captures an image of my camera, my face for example, is done a > > Trainning on the NEURAL NETWORKS and ready, it's disponible for recogzine my > > Face in all angles... > > I suggested to put this Software in a EMBEDDED SYSTEM, the problem is that, > > the C++ Software take +-5 minutes to recognize the image... :( in a 266Mhz > > computer...) > > Well, i suggested to use a EMBEDDED TRANPUTERS for Parallel Mathematical > > Computing...We intend to use a Hardware that take +-2 second for > > recognize... > > > > What do you, PICLIST USERS, think about? > > > > Miguel Wisintainer > > Master Computation

2000\04\27@154837 by Dan Michaels

Miguel wrote: ........ >My friend developted a C++ program that recognizes HUMAN FACES...He have >tested in many angles the FACE and worked very fine...ERRORS: 0%. He uses >NEURAL NETWORKS... .... >I suggested to put this Software in a EMBEDDED SYSTEM, the problem is that, >the C++ Software take +-5 minutes to recognize the image... :( in a 266Mhz >computer...) ........ You might check out various "DSP chips". They are specifically intended for doing multiply-accumulate operations, like the neural net calculations. They come in fixed point and floating point varieties. The newest TI chips run 1000-2000 MIPS: http://www.ti.com/sc/docs/products/dsp/overview.htm

2000\04\27@160735 by M. Adam Davis

Dan Michaels wrote: >The newest TI chips run 1000-2000 MIPS: So that would be GIPS, right? (Gazillion instructions per second) Well, here's one for the ol' GIPer: bottom's up! -Adam

2000\04\27@180854 by WF

Are they programable in C? I'm going there now! Thanks! Miguel ----- Original Message ----- From: Dan Michaels <oricomKILLspamLYNX.SNI.NET> To: <.....PICLISTKILLspam.....MITVMA.MIT.EDU> Sent: Thursday, April 27, 2000 4:46 PM Subject: Re: [OT] TRANSPUTERS > Miguel wrote: > ........ > >My friend developted a C++ program that recognizes HUMAN FACES...He have > >tested in many angles the FACE and worked very fine...ERRORS: 0%. He uses > >NEURAL NETWORKS... > .... > >I suggested to put this Software in a EMBEDDED SYSTEM, the problem is that, > >the C++ Software take +-5 minutes to recognize the image... :( in a 266Mhz > >computer...) > ........ > > You might check out various "DSP chips". They are specifically > intended for doing multiply-accumulate operations, like the neural > net calculations. They come in fixed point and floating point > varieties. The newest TI chips run 1000-2000 MIPS: > > http://www.ti.com/sc/docs/products/dsp/overview.htm

2000\04\27@221234 by Wagner Lipnharski

Hi Miguel... my suggestion is: Use a dog instead, it can recognize people 50 meters away in half second... hehe, just kiding. Sometimes the solution for software speed is not at the processor speed, but at the way the software can be used to "learn" and "save the learning" for future use, storing patterns and pre-processed nodes in a hudge indexed database. Sometimes is pretty difficult to "see" this possibility, how to do it? what is the best way? etc... Probably the best question is: How can WE recognize a person in seconds? Remember that it can take several minutes to do the same thing if the person is (or we are) upside down. Sometimes in a crowded store or place, looking for your lost kid you don't search for faces, but for the color of the clothes the kid is using, then for the hair, then for the arms, and so on, faster things first. Sometimes the neural network system can be programmed to look for differences instead to look for similarities... who knows? Data processing solution? buy an IBM 3090 mainframe with vector processor... :) Wagner. WF wrote: {Quote hidden}> > Here can i find low cost TRANSPUTERS? > > My friend developted a C++ program that recognizes HUMAN FACES...He have > tested in many angles the FACE and worked very fine...ERRORS: 0%. He uses > NEURAL NETWORKS... > The system captures an image of my camera, my face for example, is done a > Trainning on the NEURAL NETWORKS and ready, it's disponible for recogzine my > Face in all angles... > I suggested to put this Software in a EMBEDDED SYSTEM, the problem is that, > the C++ Software take +-5 minutes to recognize the image... :( in a 266Mhz > computer...) > Well, i suggested to use a EMBEDDED TRANPUTERS for Parallel Mathematical > Computing...We intend to use a Hardware that take +-2 second for > recognize... > > What do you, PICLIST USERS, think about? > > Miguel Wisintainer > Master Computation

2000\04\28@022100 by Dan Michaels

Wagner wrote: Hi Miguel... my suggestion is: Use a dog instead, it can recognize people 50 meters away in half second... hehe, just kiding. Sometimes the ....... >to do it? what is the best way? etc... Probably the best question is: >How can WE recognize a person in seconds? Remember that it can take >several minutes to do the same thing if the person is (or we are) upside ....... Miguel, in case you haven't seen it, you might check out CMU's Face Detector Demo Online: http://www.cs.cmu.edu/~har/faces.html

2000\04\28@035255 by Alan B Pearce

>Personally I wouldn't look at PIC processor to get the job done ;-) but how about an array of PIC's? It seems to me that they could well be used for this, but then I have not looked at neural processing......

2000\04\28@105139 by Dan Michaels

Alan B Pearce wrote: >>Personally I wouldn't look at PIC processor to get the job done ;-) > >but how about an array of PIC's? It seems to me that they could well be used for >this, but then I have not looked at neural processing...... > Speaking of which, there are people who link together piles of PCs to create "supercomputer" performance - called "Beowulf Clusters" [esp for Linux guys]: http://www.beowulf.org/ I've been expecting someone to translate those ideas into the PIC world as "PICowulf". But then, from what others have said recently, Mchp would probably hop up and put the ole clampers on use of PIC in the name. Wonder if they would object to "16C62owulf"? BTW, transputer archive: http://archive.comlab.ox.ac.uk/transputer.html

2000\04\28@122507 by WF

----- Original Message ----- From: Wagner Lipnharski <EraseMEwagnerspam_OUTTakeThisOuTUSTR.NET> To: <PICLISTspam_OUTMITVMA.MIT.EDU> Sent: Thursday, April 27, 2000 11:12 PM Subject: Re: [OT] TRANSPUTERS > Hi Miguel... my suggestion is: Use a dog instead, it can recognize > people 50 meters away in half second... hehe My house has sensors connected for a ISD1420 circuit...for PLAY a DOG message :)))) >just kiding. Sometimes the > solution for software speed is not at the processor speed, but at the > way the software can be used to "learn" and "save the learning" for > future use, storing patterns and pre-processed nodes in a hudge indexed > database. Sometimes is pretty difficult to "see" this possibility, how > to do it? what is the best way? etc... Yo're correct...my friend is rethinking about the formula used...he intends to reduce for 1 minutes for recognizing... > Probably the best question is: > How can WE recognize a person in seconds? Remember that it can take > several minutes to do the same thing if the person is (or we are) upside > down. Sometimes in a crowded store or place, looking for your lost kid > you don't search for faces, but for the color of the clothes the kid is > using, then for the hair, then for the arms, and so on, faster things > first. Sometimes the neural network system can be programmed to look > for differences instead to look for similarities... who knows? Data > processing solution? buy an IBM 3090 mainframe with vector processor... He he he! NOOO, we are thinking to start a project in the UNIVERSITY for acquire the C6000 compiler...with a project, we will get a better price...many peoples (Marcelo Yamamoto, Carl, etc) sent private e-mails to me to use DSP...it works in 1,8GFLOP....well, i'm waiting for the price of TI distributor, of this IC... We did some tests with RONALD REAGAN pictures....crying...smiling....sleeping...without Hair...in all distances...near, far...and worked very fine...the problem was to wait for 5 minutes for each picture be recognized by the Software... > :) > Wagner. > > WF wrote: > > > > Here can i find low cost TRANSPUTERS? > > > > My friend developted a C++ program that recognizes HUMAN FACES...He have > > tested in many angles the FACE and worked very fine...ERRORS: 0%. He uses > > NEURAL NETWORKS... > > The system captures an image of my camera, my face for example, is done a > > Trainning on the NEURAL NETWORKS and ready, it's disponible for recogzine my > > Face in all angles... > > I suggested to put this Software in a EMBEDDED SYSTEM, the problem is that, > > the C++ Software take +-5 minutes to recognize the image... :( in a 266Mhz > > computer...) > > Well, i suggested to use a EMBEDDED TRANPUTERS for Parallel Mathematical > > Computing...We intend to use a Hardware that take +-2 second for > > recognize... > > > > What do you, PICLIST USERS, think about? > > > > Miguel Wisintainer > > Master Computation

2000\04\28@124146 by WF

425 HTTP Error Unable to connect with remote host ;( Miguel ----- Original Message ----- From: Dan Michaels <@spam@oricomKILLspamLYNX.SNI.NET> To: <KILLspamPICLISTKILLspamMITVMA.MIT.EDU> Sent: Friday, April 28, 2000 3:19 AM Subject: Re: [OT] TRANSPUTERS {Quote hidden}> Wagner wrote: > Hi Miguel... my suggestion is: Use a dog instead, it can recognize > people 50 meters away in half second... hehe, just kiding. Sometimes the > ....... > >to do it? what is the best way? etc... Probably the best question is: > >How can WE recognize a person in seconds? Remember that it can take > >several minutes to do the same thing if the person is (or we are) upside > ....... > > Miguel, in case you haven't seen it, you might check out CMU's > Face Detector Demo Online: > > http://www.cs.cmu.edu/~har/faces.html

2000\04\28@124601 by Dipperstein, Michael

Below is a link to a paper which describes a content-based image database project done at MIT called Photobook. They had very good success searching for similarities between faces, textures, or shapes. If you plan on comparing faces, you might want to see what they did. ftp://whitechapel.media.mit.edu/pub/tech-reports/TR-255.ps.Z > {Original Message removed}

2000\04\28@125221 by Alan B Pearce

>425 HTTP Error >Unable to connect with remote host worked OK for me.

2000\04\28@130715 by M. Adam Davis

Have you tried a ronald reagan mask? -Adam WF wrote: > We did some tests with RONALD REAGAN > pictures....crying...smiling....sleeping...without Hair...in all > distances...near, far...and worked very fine...the problem was to wait for 5 > minutes for each picture be recognized by the Software... >

2000\04\28@131141 by WF

He he he! Good idea! Miguel ----- Original Message ----- From: M. Adam Davis <RemoveMEadavisTakeThisOuTUBASICS.COM> To: <spamBeGonePICLISTspamBeGoneMITVMA.MIT.EDU> Sent: Friday, April 28, 2000 2:04 PM Subject: Re: [OT] TRANSPUTERS > Have you tried a ronald reagan mask? > > -Adam > > WF wrote: > > > We did some tests with RONALD REAGAN > > pictures....crying...smiling....sleeping...without Hair...in all > > distances...near, far...and worked very fine...the problem was to wait for 5 > > minutes for each picture be recognized by the Software... > >

2000\04\28@132210 by Dan Michaels

I just tried it again, Miguel and it worked ok, you might try logging onto cs.cmu.edu and doing a search on one of these 3 guys: Henry A. Rowley (TakeThisOuTharEraseMEspam_OUTcs.cmu.edu) Shumeet Baluja (RemoveMEbalujaTakeThisOuTcs.cmu.edu) Takeo Kanade (tkEraseME.....cs.cmu.edu) or http://www.cs.cmu.edu/~simonb/face_detection/ http://www.ri.cmu.edu/labs/lab_51.html ==================================== At 01:29 PM 4/28/00 -0300, you wrote: >425 HTTP Error >Unable to connect with remote host > >;( > >Miguel > >{Original Message removed}

2000\04\30@060816 by paulb

WF wrote: > We did some tests with RONALD REAGAN > pictures....crying...smiling....sleeping...without Hair...in all > distances...near, far...and worked very fine...the problem was to wait > for 5 minutes for each picture be recognized by the Software... Doesn't he have Alzheimer's or something? Might be better to choose someone else for faster recognition? ;-) -- Cheers, Paul B.

'[OT] TRANSPUTERS'
2000\05\21@174734 by Marc

> Hi Miguel... my suggestion is: Use a dog instead, it can recognize > people 50 meters away in half second... hehe Don't forget to socket the dog. Because of its limited recognition cycles, you might want to replace it after a while. Also, it should be acousticly shielded. A thieve could otherwise trick your system into "face recognized" mode with a simple record & replay attack on the BARK IO interface.

2000\05\22@115800 by WF

:( Miguel ----- Original Message ----- From: Marc <EraseMEmarcAARGH.FRANKEN.DE> To: <RemoveMEPICLISTEraseMEEraseMEMITVMA.MIT.EDU> Sent: Sunday, May 21, 2000 12:34 PM Subject: Re: [OT] TRANSPUTERS > > Hi Miguel... my suggestion is: Use a dog instead, it can recognize > > people 50 meters away in half second... hehe > > Don't forget to socket the dog. Because of its limited recognition cycles, > you might want to replace it after a while. > > Also, it should be acousticly shielded. A thieve could otherwise trick > your system into "face recognized" mode with a simple record & replay > attack on the BARK IO interface.

2000\05\22@122022 by Dan Michaels

At 12:51 PM 5/22/00 -0300, you wrote: >:( > >Miguel Good response, Miguel [Wagner always starts these "weird" doggie things that others feel compelled to comment on - even 2 months later <:-))))]. Did you happen to get hold of a 2000 MIPS DSP processor yet? best regards, - Dan Michaels =============== >{Original Message removed}

2000\05\22@124657 by WF

I'm using the TRIAL version Software of TEXAS CHIP...nothing tested in IC...only reading the PDF... Miguel {Original Message removed}

More... (looser matching)
- Last day of these posts
- In 2000 , 2001 only
- Today
- New search...