SIGGRAPH Asia 2026
SIA: Selective Image Analogy
One before/after pair can carry several edits. SIA lets you choose which of them transfer to a new image.
Amir Alimohammadi
Ph.D. Candidate, Computing Science
GrUVi Lab, Simon Fraser University
I'm a Ph.D. candidate in computer vision and graphics at Simon Fraser University, advised by Prof. Ali Mahdavi-Amiri. I turn pretrained generative models into controllable tools for image editing, composition, and video segmentation, with work at SIGGRAPH, SIGGRAPH Asia, and ICLR. In summer 2026 I interned at Waabi, working on world models for driving.
Journey
-
2023 – presentPh.D. Computing ScienceGrUVi Lab, Simon Fraser University
-
May – Aug 2026Research InternWaabi · Toronto
-
2022 – 2023Research InternSharif University of Technology
-
2021 – 2022Research InternVITA Lab, EPFL · Switzerland
-
2018 – 2023B.Sc. Computer EngineeringSharif University of Technology
-
2017 – 2022Math Olympiad Instructor500+ hours of number theory, algebra & combinatorics
Teaching
- SFUVisual Computing Lab I & II (CMPT 742, 743) · Intro to AI · Cyber Security
- SUTLinear Algebra · Data Structures & Algorithms · Design of Algorithms · Computer Structure · Computer Networks
Honors
- 2017Gold · 35th Iranian National Mathematical Olympiad
- 2016Silver · International Mathematics Competition, Thailand
- 2015Bronze · International Mathematics Competition, China
© 2026 Amir Alimohammadi · Updated Oct 2026
Drag the nodes · fling them · click empty spaceDrag a node · tap for details
Publications * equal contribution
SIGGRAPH 2026
HairPort: Hair Import and Transfer from Single Images
3D-aware hairstyle transfer that holds up under large pose and scale differences.
Preprint · arXiv
Griffin: Generative Reference and Layout Guided Image Composition
Training-free composition from several reference images, each placed by a layout.
SIGGRAPH 2025
Cora: Correspondence-aware Image Editing Using Few-Step Diffusion
Text-guided edits in four diffusion steps that keep appearance intact through pose changes, insertions, and background swaps.
ICLR 2025
SMITE: Segment Me In TimE
Flicker-free video segmentation at any granularity from one or a few reference images.
RA-L 2024
Toward Reliable Human Pose Forecasting with Uncertainty
Uncertainty-aware 3D pose forecasting, with the UnPOSed library for fair evaluation.
Projects
- CryptoPredictions · open-source crypto forecasting library, 250+ GitHub stars
- UnPOSed · open-source 3D human pose forecasting benchmark
- Intrinsic Diffusion · monocular depth from intermediate diffusion features, no fine-tuning