Detect 3D shapes of objects and their 3D locations from one image with USL
Detect 3D shapes of objects and their 3D locations from one image with USL
Learning 3D Object Shape and Layout without 3D Supervision
arXiv paper abstract https://arxiv.org/abs/2206.07028
arXiv PDF paper https://arxiv.org/pdf/2206.07028.pdf
Project page https://gkioxari.github.io/usl
Twitter video https://twitter.com/ak92501/status/1536962925725986816
YouTube https://www.youtube.com/watch?v=PKhGIiMuRJU
A 3D scene consists of a set of objects, each with a shape and a layout giving their position in space.
Understanding 3D scenes from 2D images is an important goal, with applications in robotics and graphics.
… in predicting 3D shape and layout from a single image, most approaches rely on 3D ground truth for training which is expensive to collect at scale.
… overcome these limitations and propose a method that learns to predict 3D shape and layout for objects without any ground truth shape or layout information: instead … rely on multi-view images with 2D supervision which can more easily be collected at scale.
… demonstrate … approach scales to large datasets of realistic images, and compares favorably to methods relying on 3D ground truth.
… where reliable 3D ground truth is not available … approach outperforms supervised approaches trained on smaller and less diverse datasets.
Stay up to date. Subscribe to my posts https://morrislee1234.wixsite.com/website/contact
Web site with my other posts by category https://morrislee1234.wixsite.com/website
LinkedIn https://www.linkedin.com/in/morris-lee-47877b7b
