ScanNet

Introduced by Dai et al. in ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes

ScanNet is an instance-level indoor RGB-D dataset that includes both 2D and 3D data. It is a collection of labeled voxels rather than points or objects. Up to now, ScanNet v2, the newest version of ScanNet, has collected 1513 annotated scans with an approximate 90% surface coverage. In the semantic segmentation task, this dataset is marked in 20 classes of annotated 3D voxelized objects.

Source: A Review of Point Cloud Semantic Segmentation

Homepage

Benchmarks

Add a new result Link an existing benchmark

Task	Dataset Variant	Best Model
Semantic Segmentation	ScanNet	PTv3 + PPT
3D Object Detection	ScanNetV2	UDeerLvic
3D Instance Segmentation	ScanNet(v2)	OneFromer3D
Semantic Segmentation	ScanNetV2	CMX
Zero-shot 3D Point Cloud Classification	ScanNetV2	OpenIns3D
3D Semantic Instance Segmentation	ScanNetV2	MTML
Panoptic Segmentation	ScanNetV2	OneFormer3D
Panoptic Segmentation	ScanNet	OneFormer3D
Scene Segmentation	ScanNet	3DMV
Unsupervised 3D Semantic Segmentation	ScanNetV2	Point-GCC+PointNet++
Depth Estimation	ScanNetV2	DELTAS
Surface Normals Estimation	ScanNetV2	Bae et al.
Depth Estimation	ScanNet	Atlas
Scene Recognition	ScanNet	SSMA
Continual Semantic Segmentation	ScanNet	LGKD
3D Semantic Instance Segmentation	ScanNetV1	SGPN
3D Reconstruction	ScanNet	Atlas
3D Instance Segmentation	ScanNet	MASC
Generalized Zero-Shot Learning	ScanNet	3DGenZ