Edward James Smith

About

I am a research scientist at Borealis AI working in Machine Learning for finance.

My interests lie in the intersection of 3D object understanding and robotics. This includes 3D object generation, and prediction from a variety of input modalities and 3D understanding for robot-object interaction.

Recent Activity

For SALE: State-Action Representation Learning for Deep Reinforcement Learning accepted to NeurIPS 2023
Uncertainty-Driven Active Vision for Implicit Scene Reconstruction preprint released 2022
Active 3D Shape Reconstruction from Vision and Touch was accepted at Neurips 2021
Submitted Frame Averaging for Invariant and Equivariant Network Design to ICRL 2021
Organized the Differentiable 3D Vision and Graphics workshop at ICCV 2021

Papers

	Scott Fujimoto, Edward J. Smith, Wei-Di Chang, Shixiang Shane Gu, Doina Precup, David Meger For SALE: State-Action Representation Learning for Deep Reinforcement Learning Paper Code Cite
	Edward J. Smith, Michal Drozdzal, Derek Nowrouzezahrai, David Meger, Adriana Romero-Soriano Uncertainty-Driven Active Vision for Implicit Scene Reconstruction Paper Code Cite
	Edward J. Smith, David Meger, Luis Pineda, Roberto Calandra, Jitendra Malik, Adriana Romero, Michal Drozdzal Active 3D Shape Reconstruction from Vision and Touch Conference on Neural Information Processing Systems (NeurIPS), 2021 Paper Code Cite
	Edward J. Smith, Roberto Calandra, Adriana Romero, Georgia Gkioxari, David Meger, Jitendra Malik, Michal Drozdzal 3D Shape Reconstruction from Vision and Touch Conference on Neural Information Processing Systems (NeurIPS), 2020 Paper Code Cite
	Edward J. Smith, Krishna Murthy Jatavallabhula (Equal-first), Jean-Francois Lafleche, Clement Fuji Tsang, Artem Rozantsev, Wenzheng Chen, Tommy Xiang, Rev Lebaredian, Sanja Fidler Kaolin: A PyTorch Library for Accelerating 3D Deep Learning Research arXiv, 2019 Paper Code Cite
	Wenzheng Chen, Edward J. Smith, Jun Gao, Huan Ling*, Jaakko Lehtinen, Alec Jacobson, Sanja Fidler Learning to Predict 3D Objects with an Interpolation-based Differentiable Renderer Conference on Neural Information Processing Systems (NeurIPS), 2019 Paper Project Page Code Cite Video
	Edward J. Smith, Scott Fujimoto, Adriana Romero, David Meger GEOMetrics: Exploiting Geometric Structure for Graph-Encoded Objects International Conference on Machine Learning (ICML), 2019 Paper Code Cite
	Edward J. Smith, Scott Fujimoto, David Meger Multi-View Silhouette and Depth Decomposition for High Resolution 3D Object Representation Conference on Neural Information Processing Systems (NIPS), 2018 Paper Code Cite Video
	Edward J. Smith, David Meger Improved Adversarial Systems for 3D Object Generation and Reconstruction Conference on Robot Learning (CoRL), 2017 Paper Code Cite