Ahoy
I am trying to project a 3D point to the screen, using the formulas defined in the Camera::worldToScreen, i.e.
vec2 Camera::worldToScreen( const vec3 &worldCoord, float screenWidth, float screenHeight ) const
{
vec4 eyeCoord = getViewMatrix() * vec4( worldCoord, 1 );
vec4 ndc = getProjectionMatrix() * eyeCoord;
ndc.x /= ndc.w;
ndc.y /= ndc.w;
ndc.z /= ndc.w;
return vec2( ( ndc.x + 1.0f ) / 2.0f * screenWidth, ( 1.0f - ( ndc.y + 1.0f ) / 2.0f ) * screenHeight );
}
I’ve setup a magenta dot to be the 2D project of the 3D position of the car shown here
As you can see, if I zoom out, the dot goes further away from the car (towards bottom edge of screen). Why does this happen? I.e. It’s behaving as if the car is moving down, when that isn’t the case.
Note: I should mention – this Q is more about 2D/3D geometry than it is about cinder. I just happen to be using cinder’s “implementation” of projection, but I don’t understand the behavior I am getting from it
How are you drawing the dot? Are you setting back to an orthographic projection beforehand? gl::setMatricesWindow() for example?
I just threw together a minimal example and it worked fine for me, so is there anything special you’re doing to place the car that wouldn’t be taken into consideration by Camera::worldToScreen?
class ForumTestApp : public ci::app::App
{
public:
void setup ( ) override
{
_cam = CameraPersp ( getWindowWidth(), getWindowHeight(), 60.0f, 0.1f, 1000.0f );
_cam.lookAt( vec3 ( 0, 0, 5 ), vec3 ( 0 ) );
_camUi = CameraUi ( &_cam, getWindow() );
_camUi.setMinimumPivotDistance ( 0.0f );
}
void draw ( ) override
{
static vec3 kCubePos { 0.6, 0.3, 0.1 };
gl::clear( Colorf::gray ( 0.2f ) );
gl::setMatrices ( _cam );
{
gl::ScopedDepth depth { true };
gl::ScopedGlslProg shader { gl::getStockShader( gl::ShaderDef().lambert() ) };
gl::drawColorCube ( kCubePos, vec3(1) );
}
{
vec2 p = _cam.worldToScreen( kCubePos, getWindowWidth(), getWindowHeight() );
gl::setMatricesWindow ( getWindowSize() );
gl::ScopedGlslProg shader { gl::getStockShader( gl::ShaderDef().color() ) };
gl::ScopedColor color { Colorf ( 1, 0, 0 ) };
gl::drawSolidRect( { p - vec2(4), p + vec2(4) } );
}
}
CameraPersp _cam;
CameraUi _camUi;
};
CINDER_APP ( ForumTestApp, RendererGl );
Much thanks for the reply and example !
How are you drawing the dot? Are you setting back to an orthographic projection beforehand? gl::setMatricesWindow() for example?
I am rendering the dot using a custom 2D gui overlay (nuklear GUI), so no 3D rendering is taking part in that. Maybe it’s because the gl::setMatricesWindow() then (i wasn’t using that type of func before)
I was just wondering because the vec2 returned by worldToScreen is in 0..screenSize but perhaps nuklear renders in bottom up NDC or something. Do you get correct looking results just printing to stdout? I guess the next bit to work out is if it’s rendering or maths related.
I was just wondering because the vec2 returned by worldToScreen is in 0..screenSize
Yeah, nuklear also follows that convention – (0, 0) is top left corner, +x from left to right, and +y from top to bottom.
Do you get correct looking results just printing to stdout ?
Essentially I’m just feeding in the output of worldToScreen implementation directly to a nuklear render circle function. So any printed cout/stdout will be the same values passed into the nuklear func.
FYI, the nuklear function is called like this:
nk_fill_arc(canvas, carpos2d.x, carpos2d.y, 10.f, 0.f, 3.142f*2.f, nk_rgba(255, 0, 123, 255));
Where carpos2d represents the calculated screen pos.
It might be an issue on my side somewhere. Maybe I’m messing up certain calcs. I’ll have to check it again. Thanks for the answers !
I’ve fixed it now, I think it was my mistake when reading how to implement the worldToScreen function. All is working as expected now!