I am a PhD candidate in Computer Science. I work with Prof. Crandall in the Vision Lab. My research interest is on applying Deep Learning on vision and audio problems. Specifically I am working with Graph-Convolutional and Convolutional Neural Networks for 3D Reconstruction and Hand and Object Pose Estimation.

I spent an amazing summer at Google as a Research Intern. I worked with Tao Dong on designing an Automatic Design Analyzer to detect Material Design components using app view hierarchy and deep convolutional neural networks. We applied this design analyzer to more than 10,000 android apps to study how Material Design is used among apps with different categories, average rating and number of installs.

I also spent 6 wonderful months at Facebook Reality Labs as a research intern during Winter, Spring and Summer 2019 as a research intern. I was working on high-fidelity 3D reconstruction.

I spent Fall 2019 in Google as a research intern. I was working on detecting the social interactions in an image with multiple people in it.

I also host Koron Podcast, a podcast in Persian about Persian music.

Curriculum vitae

Work Experience

Google AI
Research Intern
Seattle, WA
Sep 2019 - Dec 2019
Facebook Reality Labs
Research Intern
Redmond, WA
Jan 2019 - Jul 2019
Google Research
Research Intern
Mountain View, CA
May 2017 - Aug 2017


C-SL: Contrastive Sound Localization with Inertial-Acoustic Sensors
Majid Mirbagheri, Bardia Doosti
[pdf] [BibTeX]
HOPE-Net: A Graph-based Model for Hand-Object Pose Estimation
Bardia Doosti, Shujon Naha, Majid Mirbagheri, David Crandall
CVPR 2020
[www] [pdf] [BibTeX] [code]
Hand Pose Estimation: A Survey
Bardia Doosti
[pdf] [BibTeX]
Observing Pianist Accuracy and Form with Computer Vision
Jangwon Lee, Bardia Doosti, David Cartledge, David Crandall, Christopher Raphael
WACV 2019
[pdf] [BibTeX]
A Computational Method for Evaluating UI Patterns
Bardia Doosti, Tao Dong, Biplab Deka, Jeffrey Nichols
[www] [pdf] [BibTeX]
A Deep Study into the History of Web Design
Bardia Doosti, David Crandall, Norman Makoto Su
WebSci 2017
[www] [pdf] [BibTeX] [code]