Low-poly 3D models

Hi! I’ve been working on some low-poly 3D stuff lately for various projects. None of these are related to Roblox. I thought it would be interesting to share this with you guys.

Not all of these models are finished, but the phase of not-knowing-what-it-is has already passed. :stuck_out_tongue:

The Voice chair: Chairtest2 - 3D model by Youri_Hoek [825cc30] - Sketchfab

Umbrella: Umbrella - 3D model by Youri_Hoek [109a1db] - Sketchfab

Old tire: Old tire - 3D model by Youri_Hoek [f35ee82] - Sketchfab

Wooden barrel: Barrel - 3D model by Youri_Hoek [c8d115a] - Sketchfab

Wooden crate: Crate1 - 3D model by Youri_Hoek [132ac74] - Sketchfab

Rusty Hotel neon sign: Neonsign1 - 3D model by Youri_Hoek [f9d710c] - Sketchfab

Roadblock 1: Roadblock3 - 3D model by Youri_Hoek [ba0e5ee] - Sketchfab

Roadblock 2: Roadblock2 - 3D model by Youri_Hoek [1eced44] - Sketchfab

2 Likes

These look great, but if you’re aiming for a low amount of polygons then why haven’t you deleted the interior faces?

Also, oops – forgot bottom:

[quote] These look great, but if you’re aiming for a low amount of polygons then why haven’t you deleted the interior faces?

Also, oops – forgot bottom:

[/quote]

That’s a good one. I think Sketchfab automatically doubles the polygons and shows them on the opposite side aswell. The models look fine in 3ds Max and engines like Unreal Engine and Unity. The downside of the second Roadblock was removed on purpose as the side never showed in the project it was for.

Would be nice to have imported meshes like these on ROBLOX. [sub]Not saying anything, just saying.[/sub]

It doesn’t double the polygons. It’s the same polygon, same texture, same data. The rendering engine in that website doesn’t factor in the normal of the polygon before drawing it.

What they should be doing is:

if Normal dot (CameraPosition - PointOnPoly) < 0 then draw it

BUT they don’t.

Nice drawing.

As a side note, this does happen in most rendering engines, but there can actually be two polygons occupying the same space with inverted normals, and in that case there is no way to know that you don’t have to iterate over the one because the other is rendered. (You’ll still hide it of course, since the normal isn’t pointing anywhere at the viewer, but it will still be less efficient than having just one polygon, since you are performing a dot operation for the one that isn’t shown.)

What most engines do allow is that you can have a polygon with two faces, this is usually a flag you can set on the polygon itself, and then it will show a face on both sides without the need of two separate polygons. I think this is the case for the example model by OP, but I’m not entirely sure.