Weakly supervised learning of multi-object 3D scene decompositions using deep shape priors
作者：
Highlights：
• Learn 3d multi-object scene decompositions for single RGB images.
• Assumes scenes with planar background and trained on synthetic images.
• Weakly supervised by using deep shape priors and differentiable rendering.
• 3D properties like shape, texture and pose are inferred for each object.
• Evaluation show generative capabilities and benefits of explicit 3D representation.
摘要
•Learn 3d multi-object scene decompositions for single RGB images.•Assumes scenes with planar background and trained on synthetic images.•Weakly supervised by using deep shape priors and differentiable rendering.•3D properties like shape, texture and pose are inferred for each object.•Evaluation show generative capabilities and benefits of explicit 3D representation.
论文关键词：Multi-object 3D scene representation learning
论文评审过程：Received 27 October 2021, Revised 17 April 2022, Accepted 19 April 2022, Available online 26 April 2022, Version of Record 10 May 2022.
论文官网地址：https://doi.org/10.1016/j.cviu.2022.103440
