Interface to the Kinect device.

See the package archive for example programs.

First you need to initalize a context. Example:

do context <- newContext
   initalize context
   ...
   shutdown context

Rather than messing around with this, it's better if you just use withContext, which does this for you:

withContext $ context -> do
  ...

All stuff with this library works within a context.

You need to select which sub devices you want to use from the Kinect (e.g. camera, motor, audio):

selectSubdevices context [Camera,Motor]

Then you open a device context through which you can control the sub devices.

withDevice context 0 $ device -> do
  ...

The second argument is which Kinect to use. You can get a count of these using

deviceCount <- countDevices context

Then you should set the depth mode you want:

setDepthMode device Medium ElevenBit

This should come before the next part, which is setting the callback:

setDepthCallback device $ payload timestamp -> do
  printf "Payload: %sn" (take 100 $ show payload)

Important: Based on the depth mode set earlier, setDepthCallback knows how to copy the payload into a vector for the callback. This is why it should come first. Arguably in future APIs a device should not be initializable without a depth mode.

Once that's done, you start the depth stream:

startDepth device

Likewise, you can grab video frames. Once you have a context, set the video mode you want using

setVideoMode device Medium RGB

In this example, we set medium resolution (640x480) with raw RGB24 Bytes.

Next, set the video callback:

setVideoCallback device $ payload timestamp -> do
  printf "Payload: %sn" (take 100 $ show payload)

Note that unlike depth, which comes in as vector of Word16's, video is a vector of Word8's.

Lastly, start the video stream:

startVideo device

Finally you need a way to receieve data. You call processEvents like this, for example:

forever $ do
  processEvents context

Calls processEvents to trigger the depth and/or video callback. Continue calling it sequentially as much as you want, but not from within the depth or video callbacks.