From the Architecture User Side

Hi.
I impress with this project. Always, in the architecture industry, we spent a lot of time to modeling tons of furniture to create the visualize image for our customers.

still, the AI is yet not perfect. there are too many polygons and faces which is not good for consumer size hardware. but only a few clicks can change some captured image into 3D model + texture is very helpful for us.

I hope that in the future you can make it better by

1. reducing polygons for flat surface (some simple model can reach 15,000 polygons with no reason). in 3D model industry we don't need mega polygon model. we spent time and use complex technique to make it light weight yet still look realistic. flat area can be textured by normal map or bump map instead of making it textured by polygon.
2. most of the captured images will comes with light (if you can get rid of light from texture this will make it fit for new environments)
3. sometime furniture come with more than 1 texture and object (fabric + wood in sofa is typically found) we need your help to separate many textures or even grouping like legs, pillows to make it easy to edit or put some reflection after all

The issue is user feedback and names no files, tests, or entry points. First review the repository's 3D generation and texturing entry points, then ask maintainers to scope one request—polygon reduction, lighting removal, or multi-texture separation—with acceptance criteria and a relevant test or example.