Noam Aigerman (pronounced "Noh-uhm", like Noah with an M at the end.)


I am a computer scientist working on problems related to machine learning and 3D geometry. My research lies at the intersection of geometry processing, computer graphics, deep learning, and optimization. Currently, I mostly focus on using geometry processing to devise thoeretically-grounded machine learning approaches for 3D problems; and, vice-versa, approaching geometry processing tasks from a machine learning perspective.
Yes, this is I!

Teaching

Winter 2024: IFT6095 - Neural Geometry Processing


Publications   (hover over a project's image for a one-sentence summary)

Top: point clouds are generated through a pretrained generative model, and are used to deform and swap meshes (bottom).
Explorable Mesh Deformation Subspaces from Unstructured 3D Generative Models
Arman Maesumi, Paul Guerrero, Vladimir G. Kim, Matthew Fisher, Siddhartha Chaudhuri, Noam Aigerman, Daniel Ritchie
ACM SIGGRAPH Asia 2023
The cow is deformed into two desired shapes described by a text prompt, in this case a turtle and a stag.
TextDeformer: Geometry Manipulation using Text Guidance
William Gao, Noam Aigerman, Thibault Groueix, Vladimir G. Kim, Rana Hanocka
ACM SIGGRAPH 2023
Face meshes from various datasets, with different triangulations, can be automatically deformed by our network into two facial expressions.
Neural Face Rigging for Animating and Retargeting Facial Meshes in the Wild
Dafei Qin, Jun Saito, Noam Aigerman, Thibault Groueix, Taku Komura
ACM SIGGRAPH 2023
A coarse mesh is progressively refined using a neural network, given additional details transmitted.
Neural Progressive Meshes
Yun-Chun Chen, Vladimir G. Kim, Noam Aigerman, Alec Jacobson
ACM SIGGRAPH 2023
A point is selected on the mesh by a user, and a region containing this point is selected by our network, such that the region can UV-mapped with low-distortion (UV map visualized with a square texture on the selected region).
DA Wand: Distortion-Aware Selection using Neural Mesh Parameterization
Richard Liu, Noam Aigerman, Vladimir G. Kim, Rana Hanocka
CVPR 2023
Top: our previous method can produce a 1-to-1 UV map of the bunny into the cross, but has distortion and scales some parts of mesh (in red). Bottom: this paper can also compute a 1-to-1 UV map, but also ensure it has low distortion and less scaling.
Isometric Energies for Recovering Injectivity in Constrained Mapping
Xingyi Du, Danny M. Kaufman, Qingnan Zhou, Shahar Z. Kovalsky, Yajie Yan, Noam Aigerman, Tao Ju
ACM SIGGRAPH Asia 2022
A chair with parts of it missing (left) is completed by our method to a full chair (middle) by copying and transforming existing parts to missing areas, while competing generative approaches produce fuzzy approximiations (right).
PatchRD: Detail-Preserving Shape Completion by Learning Patch Retrieval and Deformation
Bo Sun, Vladimir G. Kim, Noam Aigerman, Qixing Huang, Siddhartha Chaudhuri
ECCV 2022
Our method produces an atlas with a complex topology for a set of shapes.
Learning Joint Surface Atlases
Theo Deprelle, Thibault Groueix, Noam Aigerman, Vladimir G. Kim, Mathieu Aubry
ECCV Workshop on Learning to Generate 3D Shapes and Scenes, 2022
left: a UV map predicted by the network, almost identical to the ground-truth. Right: the network correctly reposes the bunny to the poses demonstrated by the human.
Neural Jacobian Fields: Learning Intrinsic Mappings of Arbitrary Meshes
Noam Aigerman, Kunal Gupta, Vladimir G. Kim, Siddhartha Chaudhuri, Jun Saito, Thibault Groueix
ACM SIGGRAPH 2022 (journal track)
The framework learns to represent a surface map from a coarse atlasnet-like MLP composed with a CNN that adds details
Neural Convolutional Surfaces
Luca Morreale, Noam Aigerman, Paul Guerrero, Vladimir G. Kim, Niloy Mitra
CVPR 2022
Various poses (orange) generated by our method from a few landmark poses (gray).
GLASS: Geometric Latent Augmentation for Shape Spaces
Sanjeev Muralikrishnan, Siddhartha Chaudhuri, Noam Aigerman, Vladimir G. Kim, Matthew Fisher, Niloy Mitra
CVPR 2022
A mesh (right) generated by optimizing the alignment of its edges to the input vector field (left).
Differentiable Surface Triangulation
Marie-Julie Rakotosaona, Noam Aigerman, Niloy Mitra, Maks Ovsjanikov, Paul Guerrero
ACM SIGGRAPH ASIA 2021
A mesh (left) is paramterized to an initial parameterization (middle) with local inversions of triangles and global overlaps of the boundary, which are then alleviated through our optimization (right).
Optimizing Global Injectivity for Constrained Parameterization
Xingyi Du, Danny M. Kaufman, Qingnan Zhou, Shahar Z. Kovalsky, Yajie Yan, Noam Aigerman, Tao Ju
ACM SIGGRAPH ASIA 2021
A sequence of reconstructed surfaces using our algorithm, exhibiting good consistent correspondences between each frame in the sequence them (visualized via texture that exhibits the correspondence).
Temporally-Coherent Surface Reconstruction via Metric-Consistent Atlases
Jan Bednarik, Vladimir G. Kim, Siddhartha Chaudhuri, Shaifali Parashar, Mathieu Salzmann, Pascal Fua, Noam Aigerman
ICCV 2021
A 2D brush is swept along a spiraling trajectory (left), tracing the golden horn (right).
Swept Volumes via Spacetime Numerical Continuation
Silvia Sellán, Noam Aigerman, Alec Jacobson
ACM SIGGRAPH 2021
Coarse voxel grids (red) are refined into different types of plants (yellow), based on the input desired style (green).
DECOR-GAN: 3D Shape Detailization by Conditional Refinement
Zhiqin Chen, Vladimir G. Kim, Matthew Fisher, Noam Aigerman, Hao Zhang, Siddhartha Chaudhuri
CVPR 2021 (oral)
Two surfaces are repsented as 2D-to-3D maps via two overfitted neural networks. Since they are both differentiable,this in turn enables optimizing a surface-to-surface map (via h) in a completely differentiable manner.
Neural Surface Maps
Luca Morreale, Noam Aigerman, Vladimir G. Kim, Niloy J. Mitra
CVPR 2021
A point cloud is meshed using our neural network.
Learning Delaunay Surface Elements for Mesh Reconstruction
Marie-Julie Rakotosaona, Paul Guerrero, Noam Aigerman, Niloy J. Mitra, Maks Ovsjanikov
CVPR 2021 (oral)
The framework of our method.
Joint Learning of 3D Shape Retrieval and Deformation
Mikaela Angelina Uy, Vladimir G. Kim, Minhyuk Sung, Noam Aigerman, Siddhartha Chaudhuri, Leonidas Guibas
CVPR 2021
Our method approximates the input heightfield surface (left) by a piecewise-developable heightfield surface (right).
Developability of Heightfields via Rank Minimization
Silvia Sellán, Noam Aigerman, Alec Jacobson
ACM SIGGRAPH 2020
A coarse mesh is subidivided via a neural network, which restores natural geometric features without over-smoothing.
Neural Subdivision
Hsueh-Ti Derek Liu, Vladimir G. Kim, Siddhartha Chaudhuri, Noam Aigerman, Alec Jacobson
ACM SIGGRAPH 2020
Mapping a mesh into a non-convex domain without any inversions, yielding a globally injective map
Lifting Simplices to Find Injectivity
Xingyi Du, Noam Aigerman, Qingnan Zhou, Shahar Kovalsky, Yajie Yan, Danny M. Kaufman, Tao Ju
ACM SIGGRAPH 2020
Deforming humanoids to match example poses via a neural network, while preserving details.
Neural Cages for Detail-Preserving 3D Deformations
Wang Yifan, Noam Aigerman, Vladimir G. Kim, Siddhartha Chaudhuri, Olga Sorkine-Hornung
CVPR 2020 (oral)
Comparison of reconstruction quality of the hybrid reconstruction versus the two components of the hybrid.
Coupling Explicit and Implicit Surface Representations for Generative 3D Modeling
Omid Poursaeed, Matthew Fisher, Noam Aigerman, Vladimir G. Kim
ECCV 2020
An embedding of a mesh into a spherical orbifold, which can tile the sphere.
Spherical Orbifold Tutte Embeddings
Noam Aigerman, Shahar Kovalsky, Yaron Lipman
ACM SIGGRAPH 2017
A convolution of a filter on a spherical surface is well-defined on the surface's toric 4-cover.
Convolutional Neural Networks on Surfaces via Seamless Toric Covers
Haggai Maron, Meirav Galun, Noam Aigerman, Miri Trope, Nadav Dym, Ersin Yumer, Vladimir G. Kim, Yaron Lipman
ACM SIGGRAPH 2017
An embedding of a mesh into a hyperbolic orbifold, which can tile the Poincare-disk model of the hyperbolic plane.
Hyperbolic Orbifold Tutte Embeddings
Noam Aigerman, Yaron Lipman
ACM SIGGRAPH Asia 2016
An embedding of a mesh into a planar orbifold, which can also be used to generate seamless quads on the mesh.
Orbifold Tutte Embeddings
Noam Aigerman, Yaron Lipman
ACM SIGGRAPH Asia 2015
A large-scale bijective parametrization of a tetrahedral mesh to a ball.
Large Scale Bounded Distortion Mappings
Shahar Kovalsky, Noam Aigerman, Ronen Basri, Yaron Lipman
ACM SIGGRAPH Asia 2015
Two identical bijective maps between two surface-meshes produced for two different cut placements.
Seamless Surface Mappings
Noam Aigerman, Roi Poranne, Yaron Lipman
ACM SIGGRAPH 2015
A low distortion bijective map between two surface-meshes.
Lifted Bijections for Low Distortion Surface Mappings
Noam Aigerman, Roi Poranne, Yaron Lipman
ACM SIGGRAPH 2014
The 'most conformal' mapping of a volumetric cube, subject to repositioning its eight corners.
Controlling Singular Values with Semidefinite Programming
Shahar Kovalsky, Noam Aigerman, Ronen Basri, Yaron Lipman
ACM SIGGRAPH 2014
A bounded-distortion, globally bijective map, mapping a tetrahedral mesh to a polycube.
Injective and Bounded Distortion Mappings in 3D
Noam Aigerman, Yaron Lipman
ACM SIGGRAPH 2013
Others
A deformation of a bar.
Computational Aspects of Mappings
Tutorial given at the IGS 2016 summer school (with Shahar Kovalsky)
Slides (pdf)
Research, in a nutshell