Question about modeling and lightmaps

After reading this article, there are not clear one thing. In documentation written

Although a separate modeling has more advantages than the related modeling:

  1. used less polygons
  2. easier modeling and subsequent change
  3. easier creating UV unwrap

Can you write more why you suggest to model a single object? This particular engine?

Attached images of separated modeling. This can cause problems?

Some things can’t be avoided, like that handle shape. but for the box shape there’s no reason to have those be separate. And in that case, you’ll end up where stuff that’s blocked behind other polygons will receive shadow and the shadow will bleed over onto the visible surfaces. Also, if you connect the pieces together then you can create UV’s where there are fewer seams and that will give better lighting results.
Also, in certain cases you might have a polygon that’s too close to another polygon and there will be z-fighting when the camera gets further away because they will get put into the same depth.

here are 3 options for modeling and already in this simple model you can see the difference in the number of polygons. When a more complicated model, the difference will be in the tens and probably hundreds of thousands of polygons. Is it not possible to set the lightmap for the first 2 options with no artifacts?
And do you think it is more correct to model?
p.s. Organic modeling is clear, this relates more to hardsurface.

I’m not clear why you believe that intersecting polygons equates more polygons.

Additionally, this shape doesn’t need the polygons to go all the way through as it does. If this is a shelf of some kind, the shelves that are horizontal, could stop at the inner wall, the vertices should be flush to the surface. This isn’t about cutting down on polygons but rather eliminating artifacts in game.

When you are modeling, don’t count the polygons by eyeballing them. A polygon doesn’t get subdivided because it intersects another shape, unless you do something to it, which usually involves other steps to your modeling like creating boolean shapes.

Look at the polygon/triangles reported for your object in your modeling system. In Maya it is reported in the upper left of your screen when turned on.

Display -> Heads Up Display -> PolyCount

When you import your fbx into UE4, the triangles will match. You model in polygons, but game engines want to break that into triangles. So the reported triangles in the HUD of Maya, will match exactly the triangle you see on your model when it is imported to UE4. You can see that number when it is importing, and when you edit that model in UE4, in the upper right corner of the Details panel associated with that model.

If you are getting 254 polygons for an object like this, there is something oddly wrong in your technique, or possibly your software that you are using that it would do this. Is this possibly Sketchup?

(in other words you should be able to make that shape you have in 48 polygons, 96 triangles).

To be clear, and if I understand you correctly. Modeling the thousands of objects needed to make a game without combining the objects made of multiple parts would be paramount to madness. Intersecting polygons doesn’t create more polygons. You will in effect have a library where every object needs to be put back together in the game engine. On top of that every translation call for a new object in game affects the game engine slowing it down.

So a shelf that is made up of nine objects put back together, is making nine engine calls constantly in game for it’s position, rather than one. This quickly balloons out of control, and you should be looking for the reverse, how to eliminate calls to the game engine on an objects position by combining duplicate objects where you can to keep the fps (frame rate per second) higher.

Additionally, UVmaps will get expensive and confusing. Mapping something like a table, with two sets of UVs, one for texture map, and one for LightMapping is work enough considering the thousands of objects needed. It is already somewhat labor intensive.

The generic UVs that come when you create a cube, doesn’t mean those UVs are correct. As soon as you begin to edit your model, the UVs then are incorrect. New UVs need to be applied when a shape is changed. Putting UVs on separate parts of an object means having many many UVs for a simple object, instead of just two sets of UVs.

You need to share some of your process. It is unclear how you are working, what software is this that divides an object based on polygon intersections.

Looks for me like Blender and Booltool, or Blender and manually modeling.
But in both cases it looks like Blender to me. ^^

I work mostly in Blender. The problem that I described apparently you understood absolutely on the contrary. I write that I see no reason to modeling most objects as contiguous mesh (show on the pictures). But the documentation says that modeling as contiguous mesh more appropriate and causes less artifacts. I like to modeling in the way described, which uses fewer polygons and using meshes intersect.
Can you attach a picture of your version of modeling of this bookcase?The picture attached is the same bookcase made in 3ds max and you can see the number of polygons. I can the same thing be done in Maya…

BTW I don’t use BoolTool


Here in Maya

One point, when you are meshing more objects into each other (no weld), ao/normals could be not calculated right(handle example(so far i know)).
When you are working in painter, then you have also the problem to autoage the circle of the handle, where it intersects normally with the pot(potside).

You’re wasting polygons on the corners, you would want one edge to go from the corner to the corner on the inside, and you have a loop going all the way around the shape which isn’t needed. You can terminate loops using triangles, and reduce the poly count. And yes, that would give better lighting results, you don’t want surfaces that are covered by other objects, the surface underneath will receive shadows and bleed to the visible area which will mess up your lighting.

In var with 48 polys I don’t have intersections

I bake all textures in blender and substance designer and don’t have problems with this

It doesn’t matter, you have polygons with surfaces that are blocked by other polygons, the shadows underneath will bleed. And you need connect the vertices on the corners to the corners of the inside.

And how in this case? Its all a split objects. why in this case, it is permissible, as is the case with the bookshelf causes so many problems?

Did you read the description?

And if you notice in the results, there’s bleeding around the elements. It’s a case where they choose not to have it as good as they would like.