Here is some test data that I am working on to check out the capabilities and speed of a CMUcam5 Pixy camera.

I am using two interfaces to set the system up :-

1. USB plugging into the Pixy allowing the PixyMon Gui to monitor,adjust & capture objects (by colour).

2. A connection between the Pixy<> Arduino on the 6pin ICSP port (aka SPI (Serial Peripheral Interface))

This is the Serial data that streams out , easily parsed to locate Colour (sig-signature) and X/Y coords.

Detected 19:
  block 0: sig: 1 x: 55 y: 147 width: 15 height: 16
  block 1: sig: 1 x: 177 y: 169 width: 19 height: 13
  block 2: sig: 1 x: 103 y: 173 width: 17 height: 16
  block 3: sig: 1 x: 204 y: 143 width: 18 height: 14
  block 4: sig: 1 x: 151 y: 143 width: 18 height: 15
  block 5: sig: 1 x: 178 y: 116 width: 18 height: 15
  block 6: sig: 1 x: 77 y: 124 width: 15 height: 12
  block 7: sig: 1 x: 59 y: 52 width: 13 height: 15
  block 8: sig: 1 x: 104 y: 53 width: 13 height: 9
  block 9: sig: 1 x: 146 y: 48 width: 15 height: 6
  block 10: sig: 1 x: 78 y: 80 width: 14 height: 8
  block 11: sig: 1 x: 172 y: 26 width: 13 height: 5
  block 12: sig: 1 x: 195 y: 75 width: 12 height: 6
  block 13: sig: 1 x: 83 y: 32 width: 12 height: 7
  block 14: sig: 4 x: 202 y: 116 width: 17 height: 17
  block 15: sig: 4 x: 177 y: 143 width: 18 height: 17
  block 16: sig: 4 x: 151 y: 169 width: 18 height: 16
  block 17: sig: 4 x: 203 y: 167 width: 18 height: 15
  block 18: sig: 4 x: 55 y: 76 width: 14 height: 14

Here is the data plotted (had to check to see if it was correct, Bingo/Bongo albeit with rotation ) :-

The Pixy processes an entire 640x400 image frame every 1/50th of a second (20 milliseconds).

So far its relatively straight forward, though lighting conditions have to be pretty constant in the example shown above. (The Pixy is going to be used for a board game (computer verses human) my Son is writing software for (in Java) , as a programming exercise to help with I.T. algorithm studies at the ETH Zurich).

Ray.Edgley

6 years 10 months ago

Hello Gareth,

Can the Pixy pro programed to recognise a generic face?
The device appears to be low power consumption, so the thought was, if it could recognise a human face, the OS could start the more power intensive face recogniser software to work out who the face is, and remain at low power using the pixy untill a face is seen.

 

Thats a shame... :-(

The only other low power solution I have found so far, and not yet fully tested is by using the Movidius NCS.
The down side of that is not all platforms are supported and it hasn't yet been borged on to MRL.

The former being the biggest problem.
Although the source code is available and one of our top developers may be able to port it to Windows and Mac so all our users can use the device.

Still the pixy has a lot of potential in smaller robotics project and can be used devices as small at an Arduino Uno...
I did find it readily available via Ebay ranging in price from AU$95 - to AU$150
Not bad value for money for something you can use in your robotics project using an Arduino as the brain.

 

kyle.clinton

6 years 10 months ago

When I unboxed my camera and and connected it looks like I have a clump of pixels that are dead or possibly there is something on the inside of the lens. I have cleaned the outside of the lens and that seems to be "clean". Have you heard of or seen anything like this?

 

Thanks

 

Kyle 

At first I though I had something on the len... however it proved to be point source light reflection, on checking the pixels all OK.

Check in with the supplier, the camera is not cheap, I would expect "Zero" dead pixels especially for its purpose in life.