3D Semantic Parsing of Large-Scale Indoor Spaces

By Iro Armeni, Ozan Sener, Amir Zamir, Helen Jiang, Ioannis K. Brilakis, Martin Fischer, Silvio Savarese. 2016

2,075 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition, Environmental Engineering, Geology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computer Vision and Pattern Recognition

In Environmental Engineering

In Geology

Citations in the corpus

PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation (2016, methodology): “We experiment on the Stanford 3D semantic parsing data set [1].”
Hypersim: A Photorealistic Synthetic Dataset for Holistic Indoor Scene Understanding (2020, background): “…the color image. alistic synthetic datasets and interactive simulation environments that have spurred rapid progress towards the goal of holistic indoor scene understanding [6, 7, 9, 10, 14, 15, 19, 22, 24, 31, 33, 36, 37, 39, 43, 44, 45, 46, 49, 52, 61, 62, 63, 65, 70, 72, 75, 76, 79, 83, 84].”
ScanNet: Richly-Annotated 3D Reconstructions of Indoor Scenes (2017, methodology): “NYU v2 [58] 464 scans 1449 frames 2D LabelMe-style [69] none some [25] TUM [81] 47 scans none aligned poses (Vicon) no SUN 3D [92] 415 scans 8 scans 2D polygons aligned poses [92] no SUN RGB-D [75] 10k frames 10k frames 2D polygons + bounding boxes aligned poses [92] no BuildingParser [3] 265 rooms ”
Dynamic Graph CNN for Learning on Point Clouds (2018, methodology): “o RGB color channels. For each pair: our prediction (left) and ground truth (right). 5.5. Indoor Scene Segmentation Data We evaluate our model on Stanford Large-Scale 3D Indoor Spaces Dataset (S3DIS) [1] for a semantic scene segmentation task. This dataset includes 3D scan point clouds for 6 indoor ”
Cognitive Mapping and Planning for Visual Navigation (2017, methodology): “ entirely in simulation. We used the ‘CMP [+6 MP3D Env]’ policy that was trained with the six additional large environments from Matterport3D dataset [14] (on top of the 4 environments from the S3DIS [7] dataset). Apart from improves performance in simulation (SPL from 59:6% to 70:8%), it also exhib”
Habitat: A Platform for Embodied AI Research (2019, background): “The availability of large-scale 3D scene datasets [5, 24, 8] and community interest in active vision tasks led to a recent surge of work that resulted in the development of a variety of simulation platforms for indoor environments [17, 7, 13, 22, 26, 3, 27, 28, 21].”
Matterport3D: Learning from RGB-D Data in Indoor Environments (2017, methodology): “Previous RGB-D datasets have provided data for single rooms or small sets of adjacent rooms [44, 7], or single ﬂoors of a building [2].”
Joint 2D-3D-Semantic Data for Indoor Scene Understanding (2017, methodology): “Recently, [16] offered water-tight mesh models of the reconstructed scenes and [1] presented a 3D point cloud dataset of large-scale indoor spaces.”
Building Generalizable Agents with a Realistic and Rich 3D Environment (2018, background): “Alternative options, yet smaller in size, include Al2-THOR (Zhu et al., 2017), SceneNet RGB-D (McCormac et al., 2017), Stanford 3D (Armeni et al., 2016) and Matterport 3D (Chang et al., 2017).”
The Replica Dataset: A Digital Replica of Indoor Spaces (2019, methodology): “The Stanford 2D3D-S dataset [4] contains 6 large-scale reconstructions with a total of 270 rooms.”
Learning Object Bounding Boxes for 3D Instance Segmentation on Point Clouds (2019, methodology): “Table 2: Instance segmentation re-sults on S3DIS dataset. mPrec mRec PartNet [32] 56.4 43.4 ASIS [51] 63.6 47.5 3D-BoNet (Ours) 65.6 47.6 We further evaluate the semantic instance segmentation of our framework on S3DIS [1], which consists of 3D complete scans from 271 rooms belonging to 6 large area”
SGPN: Similarity Group Proposal Network for 3D Point Cloud Instance Segmentation (2017, methodology): “The difference in evaluation metrics between our method and [1] is that the IoU threshold of [1] is 0.”
Deep Visual Teach and Repeat on Path Networks (2018, methodology): “Figure 4: Top view of “Area 1” in the 2D-3D-S dataset [2, 3] with paths drawn for the first 4 visual paths.”
A Review on Deep Learning Techniques Applied to Semantic Segmentation (2017, methodology): “• Stanford 2D-3D-S [51]20: is a multi-modal and largescale indoor spaces dataset extending the Stanford 3D Semantic Parsing work [64].”
InLoc: Indoor Visual Localization with Dense Matching and View Synthesis (2019, background): “With the increased availability of laser range scanners and time-of-flight (ToF) sensors, several datasets include depth data besides RGB images [4], [9],”
A Behavioral Approach to Visual Navigation with Graph Localization Networks (2019, methodology): “We perform experiments with this robot on environments from the Stanford 2D3D-S dataset [3, 4].”
Large-Scale Point Cloud Semantic Segmentation with Superpoint Graphs (2017, methodology): “ECC-VV is used for Semantic3D (there are only 15 point clouds even though the amount of points is large), while ECC-MV is used for S3DIS (large number of point clouds).”