OASIS: A Large-Scale Dataset for Single Image 3D in the Wild

Chen, Weifeng; Qian, Shengyi; Fan, David D.; Kojima, Noriyuki; Hamilton, Max; Deng, Jia

doi:10.1109/cvpr42600.2020.00076

Cited by 37 publications

(28 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We envision that detecting depth and occlusion relationships between objects can facilitate amodal tasks and vice versa. Finally, single-view 3D object detection [62,14,66,58,6,45,44,40,32,22,1,5] is related but requires labor-intensive data collection, limiting existing work to mainly indoor and self-driving environments.…”

Section: Related Datasets and Workmentioning

confidence: 99%

2.5D Visual Relationship Detection

Su¹,

Changpinyo²,

Chen³

et al. 2021

Preprint

View full text Add to dashboard Cite

Visual 2.5D perception involves understanding the semantics and geometry of a scene through reasoning about object relationships with respect to the viewer in an environment. However, existing works in visual recognition primarily focus on the semantics. To bridge this gap, we study 2.5D visual relationship detection (2.5VRD), in which the goal is to jointly detect objects and predict their relative depth and occlusion relationships. Unlike general VRD, 2.5VRD is egocentric, using the camera's viewpoint as a common reference for all 2.5D relationships. Unlike depth estimation, 2.5VRD is object-centric and not only focuses on depth. To enable progress on this task, we create a new dataset consisting of 220k human-annotated 2.5D relationships among 512K objects from 11K images. We analyze this dataset and conduct extensive experiments including benchmarking multiple state-of-the-art VRD models on this task. Our results show that existing models largely rely on semantic cues and simple heuristics to solve 2.5VRD, motivating further research on models for 2.5D perception. The new dataset is available at https://github.com/ google-research-datasets/2.5vrd.

show abstract

Section: Related Datasets and Workmentioning

confidence: 99%

2.5D Visual Relationship Detection

Su¹,

Changpinyo²,

Chen³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Planarity is also very important for various downstream applications, especially for indoor 3D reconstruction. Finally, to overcome resolution [4] and focal length variations [8], recent perspective depth estimation works resort to nearest-neighbor 3D metrics. Direct Depth Metrics.…”

Section: Metricsmentioning

confidence: 99%

Pano3D: A Holistic Benchmark and a Solid Baseline for $360^o$ Depth Estimation

Albanis,

Zioulis,

Drakoulis

et al. 2021

Preprint

View full text Add to dashboard Cite

Pano3D is a new benchmark for depth estimation from spherical panoramas. It aims to assess performance across all depth estimation traits, the primary direct depth estimation performance targeting precision and accuracy, and also the secondary traits, boundary preservation and smoothness. Moreover, Pano3D moves beyond typical intra-dataset evaluation to inter-dataset performance assessment. By disentangling the capacity to generalize in unseen data into different test splits, Pano3D represents a holistic benchmark for 360 o depth estimation. We use it as a basis for an extended analysis seeking to offer insights into classical choices for depth estimation. This results into a solid baseline for panoramic depth that followup works can built upon to steer future progress.

show abstract

“…In contrast, our approach aims at 3D understanding for dynamic objects. 3D Datasets: A number of 3D datasets have been introduced in recent years [2,3,4,7,8,14,20,28,36,39,46,48,62,66,69,71,82,83,84], all of which are commonly used for training, testing and evaluating 3D reconstruction algorithms. We summarize the properties of several common 3D datasets in Tab.…”

Section: Related Workmentioning

confidence: 99%

“…1. Datasets annotated on real world images [8,39,71,83,84] or videos [14,20] are often small compared to synthetic datasets such as SUNCG [69]. Moreover, KITTI [20] and OASIS [8] contain only modal annotations, i.e., only visible regions are annotated.…”

Section: Related Workmentioning

confidence: 99%

“…Comparisons of SAIL-VOS 3D and other 3D datasets. ShapeNet [7] ModelNet [82] PartNet [48] SUNCG [69] IKEA [39] Pix3D [71] ScanNet [14] PASCAL3D+ [84] ObjectNet3D [83] KITTI [20] OASIS [8] SAIL-VOS 3D sical 3D reconstruction focuses on finding the 3D shape of observed object parts. Nonetheless it is important for methods to predict occluded parts of objects or to infer object extent beyond the observed view [53].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation