To choose from orthographic/perspective camera for mapping object to a pixel position and scaling of that object

[nitinrhr183]

In my project, I am locating the ear points and the neck point and I am projecting 3d objects there dynamically....The video and canvas are of 640*480 merged as you can see in the video attached...

<Canvas
id="canvid"
orthographic
style={{
width: 640,
height: 480,
}}
pixelRatio={undefined}
camera={{ position: [0, 0, 350] }}
>

I have currently used orthographic camera(not sure that is the appropriate one though)...
So here I get the video pixel points(for ear and neck) as (videox,videoy) and i mapped to world coordinates as below..

const transform = (videox, videoy) => {
const k = videox - 640 / 2;
const l = videoy - 480 / 2;
//console.log("k,l=", k, l);
return [-k, -l, 10]; // position passed to render object
};
..
Now the problem I face is the appropriate scaling of object is I move back and forth wrt to webcam.When my face is near to webcam,the object shoulld enlarge and when i go far,the object should have less size....How could I do that and should I stick with orthographic for this app..

NB:For both ortho and persp,i am able to map object to the required pixel point using the same transform() function....Now the problem is back and forth scaling...

React.App.-.Google.Chrome.2021-01-13.00-36-27.mp4

[Code47X]

Have you tried using a perspective camera? Orthographic cameras don't handle depth like that. Just brainstorming here, but it seems like your transform function is not changing the depth (z) position. Its always 10, so the object would always be the same scale, even with a perspective camera. Maybe try to calculate the depth based on the distance between the ear points and set that as the objects z position with a perspective camera.

[nitinrhr183]

With perspective camera, i could locate that object to the pixel point using same transform function....The only difference was in camera z...I changed it to 320......So i should go ahead with perspective?.....how to find the correlation between depth and distance between ear points...?

[nitinrhr183]

return [-k, -l, 10]; ..If u mean that i need to change only this z position/depth ,is it possible with orthographic also...A theoretical doubt

[AndrewPrifer]

Indeed, you’ll need to use a perspective camera. Aside from the necklace looking more realistic, you can set varying positions for the necklace on the z axis, and it’ll appear smaller and bigger. You’ll need to find the factor between the depth and the distance of the ears yourself.

[nitinrhr183]

Changing z positions for necklace is actually getting it displaced out from the actual point where it is intended to project......Why is that happening