Skip to content

Dataset · huanngzh

3D-Front indoor 3D scenes with multi-view renders

huanngzh/3D-Front

Prepared 3D scenes with point clouds, depth, and normal maps for image-to-3D reconstruction tasks.

What's inside

A processed version of the 3D-Front dataset used in the MIDI-3D project. Each scene includes 3D models in .glb format, point clouds in .npy, as well as multi-view rendered RGB images, depth maps, normal maps, and camera parameters. Description language is English.

What you can do with it

  • Train and evaluate image-to-3D scene generation models
  • Estimate depth and reconstruct room geometry from multi-view images
  • Train algorithms for point clouds and polygonal meshes
Tasks
Image → 3d
Languages
English
Downloads this month
17.5K
Likes
40

How to load it

from datasets import load_dataset

ds = load_dataset("huanngzh/3D-Front")