screenToWorld3D

Autonomous Machines DRIVE AGX DRIVE AGX General
, ,
#1

Please provide the following info:
Hardware Platform: DRIVE AGX Pegasus™ Developer Kit
Software Version: DRIVE Software 10
Host Machine Version: native Ubuntu 18.04
SDK Manager Version: 1.4.0.7363

In render_engine sample, clicking on tiles shows an {x, y, z} coordinate that is generated based on the clicked screen position using dwRenderEngine_screenToWorld3D.

It seems that the world coordinate always has z=0.

What I need to do is click on a point after rendering a point cloud with DW_RENDER_ENGINE_PRIMITIVE_TYPE_POINTS_3D and get the 3D coordinates of a point close to where the mouse is clicked.

Is screenToWorld3D the correct approach, or is there a different way?

glReadPixels
#3

Hello. Any tips on this? Thanks!

#4

Bumping this again. Does the question even make sense? Thanks.

#5

Hi @gpu_power,

Below are the steps of finding the closest point. FYI.

  1. click on screen and pass pixel coords to dwRenderEngine_screenToWorld API
  2. Using the 3D point obtained from the above API, search through the lidar points and find the point that is closest to that 3D point.
#6

Thanks @VickNV

Not sure what I’m doing wrong, but dwRenderEngine_screenToWorld always results in world position of {-1, 1, 0}.

#7

I’ll check with the team and get back to you. Thanks!

#9

To make it clear, the problem is the 3D point returned by dwRenderEngine_screenToWorld3D always results in z=0. This is the same as in the render_engine example as well as my own code. The values of x and y seem correct.

What I am trying to do is commonly done in opengl by first calling glReadPixels to get the GL_DEPTH_COMPONENT (depth value) of the pixel coords. Then calling gluUnProject, with the pixel coords and depth value. This gets the 3D point {x, y, z} in opengl.

However, I can’t do the same in Driveworks. I think gluUnProject is equivalent to dwRenderEngine_screenToWorld3D, but as I described in my original post, z always results in 0.

#10

Thank you for helping to find the issue! We are fixing it and will update you here.

#11

@VickNV can you please help with a workaround?

If I can use glReadPixels to get the GL_DEPTH_COMPONENT, then I can do the matrix operations myself.

However, although glReadPixels works on the host Ubuntu PC (i.e. it returns sensible values of depth: 0.0 < depth < 1.0), it always returns 0.0 on the target Pegasus platform.

How do I get glReadPixels working on the target?

#12

Could you create another topic for it and upload the sample for our reproducing? Thanks!