# What is a good approach to reduce the number of faces on this chair model

See the model below. I first scanned it with a 3D scanner and then remodelled every part of it. I used the subdivision modifier a lot. Resulting in a smooth surface. Sometimes a surface is completely flat, like the top of the seat for instance. I used the subdiv modifier here to get a smooth curvature on the edges.

Is it good practice to reduce the number of faces on flat surfaces? I am thinking of modelling for VR environments for instance. What is a good approach to reduce the number of faces on this seat, for instance.

• Separate all the faces (object parts) you want to simplify. So you'll have another object. Set a 'decimate' modifier to it and use 'unsubdivide' option (probably with a pair amount of iterations). – lemon Jul 15 '19 at 7:43
• @lemon First splitting the object to use the modifier is a great idea! And I should join this object later on right? Preferably also connecting the 2 meshes to form a single on, right? – Mike de Klerk Jul 15 '19 at 11:16

If you were taking a sub-d approach, I don't see anything in this particular model which requires tricky transitions from high-density to low-density mesh, especially if you can model separate parts as separate parts, which is the general rule, unless you need a manifold surface (for example, for printing).

You would be aiming for the lowest possible number of edge-loops to capture the curvature of the piece .. let's say, the seat-base...

..shown here with a) the model as made, as a cage over the subdivided surface, b) the cage conforming to the subdivision modifier, optimal display on, and c) the underlying subdivision, level 2.

So, if you chose to go that way, you would dissolve nearly all the edge-loops, keeping only those needed to capture the curvature and sharpen edges, assign a Subdivision Surface modifier, and apply it at the last possible moment, if at all.

If you were really pushed for your poly-count, completely flat areas isolated from the surrounding curvature can be N-gons. You could apply the subdivision, and dissolve all the edges in the flat area (maybe insetting the N-gon to protect the surrounding curves), and cut in just enough edges to get rid of smoothing artifacts resulting from triangulation.

You can reduce the number of quads on a flat surface without generating Ngons with edges poles in red, by doing something like this (I don't know the name of that specific kind of retopology)

High poly is useful for boolean operations or deformation, to reduce the number of polygons, and keeping detail you can use for baking.