Hello there!


I am a Computer Vision Researcher at the Architecture and Artificial Intelligence Lab ( AR2IL ) working with Matias del Campo and Justin Johnson. My research interests lies in developing agents that can learn to understand the underlying representation of our 3D world and power real-world applications for steerable 3D scene generation.
I received my Master’s degree in Electrical & Computer Science from University of Michigan, Ann Arbor. During my masters, I spent my summer at Intvo working on a pedestrian intent estimation framework in the context of ego-centric vehicles. Previously, I worked at ZS Associates, where I applied the concepts of Computer Vision and NLP to innovate solutions across Healthcare, Gaming and Software industry.
Feel free to say hi: dasyed at umich dot edu
ūüĎč Offering help: Feel free to book a slot on my calendar to talk about grad school, computer vision, etc.

News


[01/2021]     Started as a Research Assistant at the Architecture & Artificial Intelligence Lab (AR2IL)
[09/2020]     Paper with Naman Gandhi, Arushi Arora & Nilesh Kadam accepted at ICMLA 2020!
[06/2020]     Spending summer at at Intvo Inc as Computer Vision Engineer Intern
[09/2019]     Started as a Masters student at the University of Michigan!

Publications


DeepGamble: Towards unlocking real-time player intelligence using multi-layer instance segmentation and attribute detection
Danish Syed*, Naman Gandhi*, Arushi Arora* and Nilesh Kadam
[* indicates equal contribution]
ICMLA 2020
[arXiv] [project]


Projects


VO_benchmark report | code | video
A comparative study of various detector-descriptor combinations used in Visual Odometry to explore the effect of semantic understanding in localization for SE(3) poses.
SPPFNet report
A particle filter based end-to-end pose estimator where each particle learns latent embedding to infer pose, object likelihood, and re-sampling objective iteratively.
SICGAN report | code
An end-to-end conditional GAN framework for generating 3D objects from single RGB image. It was able to get better qualitative 3D reconstructions as compared to the baseline.
MedLens code | demo
A Bi-Directional Attention Flow network based web app for health-care researchers which answers factual questions based on uploaded documents or searched on PubMed database

Teaching


EECS 442: Computer Vision (Winter 2021)
GSI with Justin Johnson & David Fouhey
EECS 598: Deep Learning for Computer Vision (Fall 2020)
GSI with Justin Johnson


Deepdesign: Architecture & 3D Neural Networks (Summer 2020)
Instructor with Matias del Campo & Sandra Manninger
Arch 662: Architecture & Artificial Intelligence (Winter 2020)
TA with Matias del Campo & Sandra Manninger