InteriorNet Large-Scale Indoor Scene Recognition Dataset

InteriorNet is an RGB-D dataset for large-scale indoor scene understanding and mapping. The dataset contains 20 million images created by the pipeline:

(A) The authors collected approximately one million CAD models from the world's leading furniture manufacturers.

(B) Based on these models, about 1,100 professional designers have created about 22 million interior layouts. Most of these layouts have been used in the real world.

(C) For each layout, the authors generate several configurations to simulate different lighting conditions and changing scenarios in daily life.

(D) The authors provide an interactive simulator (ViSim) to help create ground truth IMU, event, and monocular or stereo camera trajectories, including hand-drawn, random movement, and neural network-based real trajectories.

(E) All supported image sequences and ground truth.