PBR-Rooms

PBR-Rooms: A PBR-Complete Indoor Benchmark for Inverse Rendering Evaluation Dataset Summary PBR-Rooms is a synthetic indoor benchmark for evaluating multi-channel inverse rendering under physically based rendering (PBR) ground truth. It is designed to test whether models can recover both scene geometry and PBR material properties from RGB images. The complete generated PBR-Rooms corpus contains 1,000 procedurally generated indoor scenes across five room categories:… See the full description on the dataset page:

Source: Hugging Face Hub (pbr-rooms-benchmark/PBR-Rooms). Metadata imported from the dataset’s Hub tags.