ScanNet is a large-scale RGB-D dataset containing 1513 scans of real-world indoor environments, with 2.5 million RGB-D images. The dataset includes 3D surface reconstructions, semantic segmentations, and CAD model alignments. It was collected using an easy-to-use and scalable RGB-D capture system that includes automated surface reconstruction and crowdsourced semantic annotation. The dataset is valuable for 3D scene understanding tasks such as 3D object classification, semantic voxel labeling, and CAD model retrieval. ScanNet provides dense 3D reconstructions with instance-level object category labels and aligned CAD models, making it a rich resource for research in 3D scene understanding. The dataset was collected using a framework that allows untrained users to capture and annotate RGB-D data, and it includes a complete open-source acquisition and annotation framework for dense RGB-D reconstructions. ScanNet has been used to train 3D deep networks and has demonstrated state-of-the-art performance on several 3D scene understanding tasks. The dataset is larger than any other existing RGB-D dataset and includes more detailed annotations. ScanNet has been used to evaluate the performance of 3D object classification, semantic voxel labeling, and CAD model retrieval, and has shown that the dataset enables the use of deep learning methods for 3D scene understanding tasks with supervised training. The dataset has also been used to compare performance with other existing datasets and has shown that ScanNet provides better results for these tasks. The dataset includes a variety of indoor spaces such as offices, apartments, and bathrooms, and has a diverse set of scenes ranging from small to large. The dataset has been used to train and test various 3D scene understanding tasks, and has demonstrated the effectiveness of ScanNet in achieving state-of-the-art performance. The dataset is a valuable resource for researchers in the field of 3D scene understanding and has the potential to inspire future work on many other tasks.ScanNet is a large-scale RGB-D dataset containing 1513 scans of real-world indoor environments, with 2.5 million RGB-D images. The dataset includes 3D surface reconstructions, semantic segmentations, and CAD model alignments. It was collected using an easy-to-use and scalable RGB-D capture system that includes automated surface reconstruction and crowdsourced semantic annotation. The dataset is valuable for 3D scene understanding tasks such as 3D object classification, semantic voxel labeling, and CAD model retrieval. ScanNet provides dense 3D reconstructions with instance-level object category labels and aligned CAD models, making it a rich resource for research in 3D scene understanding. The dataset was collected using a framework that allows untrained users to capture and annotate RGB-D data, and it includes a complete open-source acquisition and annotation framework for dense RGB-D reconstructions. ScanNet has been used to train 3D deep networks and has demonstrated state-of-the-art performance on several 3D scene understanding tasks. The dataset is larger than any other existing RGB-D dataset and includes more detailed annotations. ScanNet has been used to evaluate the performance of 3D object classification, semantic voxel labeling, and CAD model retrieval, and has shown that the dataset enables the use of deep learning methods for 3D scene understanding tasks with supervised training. The dataset has also been used to compare performance with other existing datasets and has shown that ScanNet provides better results for these tasks. The dataset includes a variety of indoor spaces such as offices, apartments, and bathrooms, and has a diverse set of scenes ranging from small to large. The dataset has been used to train and test various 3D scene understanding tasks, and has demonstrated the effectiveness of ScanNet in achieving state-of-the-art performance. The dataset is a valuable resource for researchers in the field of 3D scene understanding and has the potential to inspire future work on many other tasks.