What question did this study set out to answer?

To introduce a large-scale, richly annotated RGB-D video dataset alongside a scalable capture pipeline to advance supervised deep learning methods for 3D indoor scene understanding.

February 14, 2017Open Access

ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes

View Full Paper

Ask AI

Bookmark

Authors

ADAngela DaiWorld Health Organization - Pakistan MSManolis SavvaSimon Fraser University MHMaciej HalberPrinceton University

Discussion

Member takes

Implication

Dataset development study demonstrates state-of-the-art 3D scene understanding via 2.5 million RGB-D views across 1513 indoor scenes, indicating the value of scalable semantic annotation.

Key Points

To introduce a large-scale, richly annotated RGB-D video dataset alongside a scalable capture pipeline to advance supervised deep learning methods for 3D indoor scene understanding.
Engineered an RGB-D capture framework combining automated surface reconstruction with crowdsourced semantic segmentation.
Acquired 2.5 million RGB-D video views across 1,513 indoor scenes labeled with 3D camera poses, surface reconstructions, and semantic annotations.
Evaluated supervised 3D scene understanding models on benchmarks including 3D object classification, semantic voxel labeling, and CAD model retrieval.
Achieved state-of-the-art benchmark performance on 3D object classification and semantic voxel labeling tasks using the annotated dataset.
Demonstrated improved performance on CAD model retrieval compared to models trained on existing, smaller RGB-D datasets.

Cite This Study

Dai et al. (2017) studied this question.

synapsesocial.com/papers/6a8315d8c009eb9b7eefe2cd https://doi.org/10.48550/arxiv.1702.04405

View Full Paper

Ask AI

Bookmark

Implication

Dataset development study demonstrates state-of-the-art 3D scene understanding via 2.5 million RGB-D views across 1513 indoor scenes, indicating the value of scalable semantic annotation.

ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes

Discussion

Member takes

Implication

Key Points

Cite This Study

ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider