Sahar Dastani

I am a PhD student at Mila/ÉTS, advised by Prof. Christian Desrosiers and Prof. Hervé Lombaert. My research sits at the intersection of computer vision, deep learning, and efficient neural networks, with a focus on robustness under distribution shifts and geometric transformations. I am also a machine learning research intern at Borealis AI.

Previously, I was a machine learning research intern at Zebra. I also worked as a research assistant at CHU Sainte-Justine in Montreal under the supervision of Dr. Philippe Jouvet. I received my M.Sc. and B.Sc. in Computer Science from Amirkabir University of Technology (AUT), under the supervision of Dr. M. Hassan Shirali-Shahreza and Dr. Mohammad Akbari.

Email Google Scholar LinkedIn GitHub

Highlights & News

01/2026 Joined Borealis AI as a Machine Learning Research Intern.
09/2025 Two papers accepted at NeurIPS 2025: TRUST and OVSS TTA.
08/2025 Served as a reviewer for NeurIPS 2025.
07/2025 Prompt Learning With Bounding Box accepted at IEEE T-BME 2025.
06/2025 Purge-Gate accepted at ICCV 2025.
05/2025 SMART-PC accepted at ICML 2025.
02/2025 Two papers accepted at CVPR 2025: Spectral VMamba and Spectral Informed Mamba.
02/2025 Reviewer for IEEE TMI-2024.
01/2025 Reviewer for CVPR 2025.
01/2025 TTA in Point Clouds with Weight Averaging accepted at WACV 2025 (Oral).
09/2024 Joined Zebra as a Machine Learning Research Intern.
09/2024 WATT accepted at NeurIPS 2024.
02/2024 Abstract on Psychomotor Agitation Detection accepted at WFPICCS.
01/2024 Awarded the Upper Bound Talent Bursary 2024.
12/2023 Awarded the Professor Cho Diversity Scholarship.
09/2023 TA for Representation Learning at UdeM with A. Agrawal.

Publications

TRUST: Test-Time Refinement using Uncertainty-Guided SSM Traverses

Sahar Dastani, A. Bahri, G. A. Vargas Hakim,

more authors

M. Yazdanpanah, M. Noori, D. Osowiechi, S. Barbeau, I. Ben Ayed, H. Lombaert, C. Desrosiers.

NeurIPS 2025.

paper code

Spectral State Space Model for Rotation-Invariant Visual Representation Learning

Sahar Dastani, A. Bahri, M. Yazdanpanah,

more authors

M. Noori, G. A. Vargas Hakim, D. Osowiechi, F. Beizaee, A. Mondal, H. Lombaert, C. Desrosiers.

CVPR 2025.

paper code

Test-Time Adaptation of Vision-Language Models for Open-Vocabulary Semantic Segmentation

D. Osowiechi, M. Noori, G. A. Vargas Hakim, A. Bahri, M. Yazdanpanah, Sahar Dastani,

more authors

F. Beizaee, I. Ben Ayed, C. Desrosiers.

NeurIPS 2025.

paper code

Prompt Learning With Bounding Box Constraints for Medical Image Segmentation

M. Gaillochet, M. Noori, Sahar Dastani,

more authors

C. Desrosiers, H. Lombaert.

IEEE T-BME 2025.

paper code

SMART-PC: Skeletal Model Adaptation for Robust Test-Time Training in Point Clouds

A. Bahri, M. Yazdanpanah, Sahar Dastani,

more authors

M. Noori, G. A. Vargas Hakim, D. Osowiechi, F. Beizaee, I. Ben Ayed, C. Desrosiers.

ICML 2025.

paper code

Purge-Gate: Efficient Backpropagation-Free Test-Time Adaptation for Point Clouds via Token Purging

M. Yazdanpanah, A. Bahri, M. Noori, Sahar Dastani,

more authors

G. A. Vargas Hakim, D. Osowiechi, I. Ben Ayed, C. Desrosiers.

ICCV 2025.

paper code

Spectral Informed Mamba for Robust Point Cloud Processing

A. Bahri, M. Yazdanpanah, M. Noori, Sahar Dastani,

more authors

M. Cheraghalikhani, D. Osowiechi, G. A. Vargas Hakim, F. Beizaee, I. Ben Ayed, C. Desrosiers.

CVPR 2025.

paper code

Test-Time Adaptation in Point Clouds: Leveraging Sampling Variation with Weight Averaging

A. Bahri, M. Yazdanpanah, M. Noori, Sahar Dastani,

more authors

M. Cheraghalikhani, D. Osowiechi, G. A. Vargas Hakim, F. Beizaee, I. Ben Ayed, C. Desrosiers.

WACV 2025. (Oral)

paper code

WATT: Weight Average Test Time Adaptation of CLIP

D. Osowiechi, M. Noori, G. A. Vargas Hakim, M. Yazdanpanah, A. Bahri, M. Cheraghalikhani, Sahar Dastani,

more authors

F. Beizaee, I. Ben Ayed, C. Desrosiers.

NeurIPS 2024.

paper code

Psychomotor agitation detection using deep learning

Sahar Dastani, A. Harakeh, P. Jouvet,

more authors

S. Kahou.

WFPICCS PCCM Journal 2024.

paper

Relational UNet for Image Segmentation

I. Sheth, P. Braga, S. Sujit, Sahar Dastani,

more authors

S. Kahou.

MICCAI Workshop 2023.

paper code

Experiences

Machine Learning Research Intern

Borealis AI · Montréal, Canada

January 2026 - April 2026

Machine Learning Research Intern

Zebra Technologies · Montréal, Canada

September 2024 - December 2025

Program Reporter

CIFAR DLRL Summer School · Montréal, Canada

July 2023

Machine Learning Researcher

Mila - Quebec AI Institute · Montréal, Canada

September 2022 - Present

Machine Learning Researcher

CHU Sainte-Justine · Montréal, Canada

September 2022 - February 2024

Machine Learning Expert

Holoo Software Engineering Company · Tehran, Iran

May 2021 - July 2021

Education

Ph.D. in Computer Science

École de technologie supérieure (ÉTS) · Montreal, Canada

Supervisors: Prof. Christian Desrosiers, Prof. Hervé Lombaert

Thesis Title: Deep Learning Approaches for Efficient Neural Network Design in Computer Vision

2022 – 2026

GPA: 4.3 / 4.3

M.Sc. in Computer Science

Amirkabir University of Technology (AUT) · Tehran, Iran

Supervisors: Dr. M. Hassan Shirali-Shahreza and Dr. Mohammad Akbari

Thesis Title: 3D Human Pose Estimation Using Neural Networks

2020 – 2022

GPA: 4.00 / 4.00

B.Sc. in Computer Science

Amirkabir University of Technology (AUT) · Tehran, Iran

2016 – 2020

GPA: 3.84 / 4.00

Contact

Email: sahar.dastani-oghani.1@ens.etsmtl.ca

Scholar: profile · LinkedIn: profile · GitHub: Sahardastani