ScanNet++ (v2)

ScanNet++ is a high-fidelity 3D indoor-scene dataset pairing sub-millimeter laser scans with high-resolution DSLR and commodity iPhone RGB-D captures, enabling joint benchmarks for novel-view synthesis and 3D semantic understanding. The v2 release (Dec 2024) expands to 1,000+ scenes and adds 360° panoramic and iPhone-NVS benchmarks (2025). From Matthias Nießner's lab; a widely used evaluation standard for 3D reconstruction and Gaussian-splatting methods.