search

diablodale / dp.kinect2

Kinect v2 (XBox One) plugin for Cycling '74 Max
21 stars 4 forks source link

Depthmap with @type long gives unexpected result #61

Closed Teufeuleu closed 1 year ago

Teufeuleu commented 1 year ago

Description

I use [dp.kinect2 @depthmap 1 @type long] in order to perform bitwise operations on the depth matrix. The main goal is to record a video I can use as in the three.js kinect webgl example, which implies encoding depth as RGB values (as shown in this modified freenect file).

However, the long matrix output seems to only be the truncated values of the float32/64 output. As a proof, a [jit.3m] gives, say 0 as min, 1.9 as mean and 7.9 as max for a float32/64 output, while the long output gives 0, 1.4 and 7 respectively. I would expect the long output to keep the same resolution as a float32/64 output, and not being a truncated version of it with the equivalent of a 3bit resolution.

Setup

Steps to reproduce

  1. Start Max 8.
  2. Open the dp.kinect2 help patch
  3. Set @depthmap 1 @type long as attributes of [dp.kinect2]
  4. Create a [jit.3m] or [jit.cellblock] and connect its input to the leftmost output of [dp/kinect2]
  5. Press the trigger to open the kinect stream
  6. Check the outputs given by [jit.3m] or [jit.cellblock], and compare those to when @type is set to float32/64.

Result

The long matrix output seems to only be the truncated values of the float32/64 output.

Expected Result

I would expect the long output to keep the same resolution as a float32/64 output, and not being a truncated version of it with the equivalent of a 3bit resolution..

Workarounds

A workaround might be to somehow translate the float64 output as a long, but I'm not sure how to do it without loss of information. All my attempts have failed so far.

Max console output with @verbose 1:

Nothing special (last 3 lines are only when opening the device, then it runs smoothly)

dp.kinect2 1.2.20210521+723f5c9 Win64, Copyright (c) 2014-2021 Dale Phurrough. This software comes with ABSOLUTELY NO WARRANTY.
-All rights reserved except for those documented in the dp.kinect2 terms at https://hidale.com and granted by purchase/agreement to any entity listed as registered below.
-Compiled with Eigen 3.3.9 licensed under Mozilla Public License, v2.0. Source code & license at http://eigen.tuxfamily.org/
-BY USING THIS SOFTWARE, YOU ACCEPT THESE TERMS. IF YOU DO NOT ACCEPT THEM, DO NOT USE THIS SOFTWARE.
dp.kinect2 registered for [name]. Valid 12 Jan 2023 - 31 Dec 2999. Sn 1680.
kinect performance caution: power setting (USB selective suspend) should be disabled
dp.kinect2: data unavailable for request
dp.kinect2: data unavailable for request
dp.kinect2: data unavailable for request
diablodale commented 1 year ago

That is (technically) expected behavior. I recommend you use @distmeter 0 so that you receive millimeters with your @type long

Depth long output is an integer and has no fractional values. Therefore its possible values when @distmeter 1 are only 0,1,2,3,4,5,6,7,8. With @distmeter 0 output is in millimeters and therefore can be the integer range 0-8000.

diablodale commented 1 year ago

Or...you could keep @distmeter 1 and output floats. Then use Max's jitter operators to multiply that by a number (like 1000) which changes 1.5 meter depth to 1500.0 mm depth. And then force the Jitter matrix into a long. It really depends on what processing you need to do after dp.kinect2 outputs the depth matrix. You have flexible choice :-)

Teufeuleu commented 1 year ago

I knew I missed some attribute. Thanks!

diablodale commented 1 year ago

No worries. I've added this scenario to the new documentation (work in progress) https://docs.hidale.com/max/attributes/distmeter/