Rel3D: A Minimally Contrastive Benchmark for Grounding Spatial Relations in 3D

12/03/2020
by   Ankit Goyal, et al.
1

Understanding spatial relations (e.g., "laptop on table") in visual input is important for both humans and robots. Existing datasets are insufficient as they lack large-scale, high-quality 3D ground truth information, which is critical for learning spatial relations. In this paper, we fill this gap by constructing Rel3D: the first large-scale, human-annotated dataset for grounding spatial relations in 3D. Rel3D enables quantifying the effectiveness of 3D information in predicting spatial relations on large-scale human data. Moreover, we propose minimally contrastive data collection – a novel crowdsourcing method for reducing dataset bias. The 3D scenes in our dataset come in minimally contrastive pairs: two scenes in a pair are almost identical, but a spatial relation holds in one and fails in the other. We empirically validate that minimally contrastive examples can diagnose issues with current relation detection models as well as lead to sample-efficient training. Code and data are available at https://github.com/princeton-vl/Rel3D.

READ FULL TEXT

page 2

page 5

page 8

page 10

08/07/2019

SpatialSense: An Adversarially Crowdsourced Benchmark for Spatial Relation Recognition

Understanding the spatial relations between objects in images is a surpr...
08/24/2021

Support-Set Based Cross-Supervision for Video Grounding

Current approaches for video grounding propose kinds of complex architec...
07/17/2020

Visual Relation Grounding in Videos

In this paper, we explore a novel task named visual Relation Grounding i...
07/24/2022

Hierarchical Semi-Supervised Contrastive Learning for Contamination-Resistant Anomaly Detection

Anomaly detection aims at identifying deviant samples from the normal da...
07/26/2020

Contrastive Visual-Linguistic Pretraining

Several multi-modality representation learning approaches such as LXMERT...
10/11/2021

Pano-AVQA: Grounded Audio-Visual Question Answering on 360^∘ Videos

360^∘ videos convey holistic views for the surroundings of a scene. It p...
03/15/2017

Learning to Discover Cross-Domain Relations with Generative Adversarial Networks

While humans easily recognize relations between data from different doma...