Sana Arastehfar

Intrinsic Mollification - SGI MIT

2023

• his project is conducted under supervision of Keenan Crane, where we investigate maintaining triangle quality in geometry processing algorithms without altering mesh connectivity. GitHub

​

Tangible NeRFs: Geometry-guided NeRF Exploration - SGI MIT

2023

• This project is conducted under supervision of Ilke Demir based on recent scene understanding and editing approaches using the combination of Diffusion Models and NeRFs. This project is based on enhancing NeRFs performance through stabilizing the output using geometry and text guidance. GitHub

​

Single Agent Behavior Prediction Using Linear Temporal Logic (LTL) in Soccer

2023

• Master’s thesis in sport analytics under supervision of professors Muise and Pfaff. The goal of this research is to predict a single soccer player's behavior by extracting the Linear Temporal Logic (LTL) formula. Taking advantage of BayesLTL, a modern tool for extracting LTL formulas in a Bayesian and contrastive framework, we extract a player’s longstanding behavior in different situations (such as before and after losing ball possession, before and after receiving a goal, and etc.) in the form of interpretable human-readable LTL formulas. GitHub

​

Visual Representation Learning of Colorectal Cancer in Histology Images Using Contrastive Learning

2022

• Implementation of Simple Contrastive Learning, Self-supervised learning with BYOL and Supervised Contrastive Learning for the first time on CRC100k dataset. The purpose of this work is to assess the effects of contrastive loss on medical datasets, specifically in a supervised setting. GitHub

​​

Smart Meter Data Analysis for Prediction of Residential Energy Consumption

2020 till 2021

• Final project of Bachelor’s degree under the supervision of Professor Mohammadreza Jabbarpour. The goal of the project is to propose a novel deep learning method for estimation and forecasting the consumption loads using smart meter data. Our proposed method consists of employing Graph Neural Network and LSTM networks to capture both temporal and spatial information. GitHub

​

Implementation of Personal Library for Classic Machine Learning Method

2019

• Implementation of classic Machine Learning methods such as Regression, KNN, Decision Trees, Naïve Bayes… in Octave/Matlab. GitHub

​

Heart Attack Prediction & Implementation of Portable ECG

2015 till 2019

• Under supervision of Professor Hamid Reza Marateb, and together with my colleagues in “Zist Abzar Pars,” a healthcare startup co-founded by me, we worked on heart attack prediction using patients Electrocardiography signal profiles. We processed and classified signals for ECG using Support Vector Machine (SVM) and k-Nearest Neighbors (k-NN) methods for arrhythmia detection. Our effort manifested itself in designing and implementing an innovative portable ECG. The ultimate achievement of this project was introducing the SnapECG wearable device, with the capability to store patients’ heart signals. ​

​

​

​