Litcius/Paper detail

DeepV2D: Video to Depth with Differentiable Structure from Motion

Zachary Teed, Jia Deng

2020King Abdullah University of Science and Technology Repository (King Abdullah University of Science and Technology)68 citations

Abstract

We propose DeepV2D, an end-to-end deep learning architecture for predicting depth from video. DeepV2D combines the representation ability of neural networks with the geometric principles governing image formation. We compose a collection of classical geometric algorithms, which are converted into trainable modules and combined into an end-to-end differentiable architecture. DeepV2D interleaves two stages: motion estimation and depth estimation. During inference, motion and depth estimation are alternated and converge to accurate depth.

Topics & Concepts

Differentiable functionComputer scienceArtificial intelligenceMotion (physics)Motion estimationRepresentation (politics)Computer visionInferenceArtificial neural networkStructure from motionImage (mathematics)MathematicsMathematical analysisLawPolitical sciencePoliticsAdvanced Vision and ImagingImage Processing Techniques and ApplicationsOptical measurement and interference techniques
DeepV2D: Video to Depth with Differentiable Structure from Motion | Litcius