Ashutosh Kumar

AN AVID TECHIE

Ph.D.
Computer Vision and AI
MLOps
Web3 and smart contract
Technophile

About me

hero

My name is Ashutosh, and I am currently a second-year Ph.D. student at the Graduate School of Engineering, University of Tokyo, Japan. Since my Master's, I have been affiliated with the Human-Centered Urban Informatics lab under the guidance ofProfessor Yoshihide Sekimoto.

My current research work is on the real-time citywide reconstruction of cross-sectional traffic flow from multiple moving cameras on lightweight edge devices.

As an intern, freelancer, and part-time employee, I have worked for various institutions, startups, and companies in the field of computer vision for the past five years. In spite of having spent most of my career in computer vision domain of deep learning, I have recently grown fascinated by the field of natural language processing after latest advancements in NLP algorithms.

All kinds of technologies captivate me, and I constantly research and learn about emerging technologies. The satisfaction I receive from working on and resolving problems with my acquired knowledge is immense.

You can contact me here: ash_k@mit.edu

Latest

Achievements/News
I have been accepted as a visiting student at the Senseable City Lab, Massachusetts Institute of Technology (MIT), where I will be working with Professor Carlo Ratti and Paolo Santi on real-time urban monitoring tasks using computer vision and deep learning.
Achievements
My fourth Ph.D. research paper titled "Vehicle re-identification and trajectory reconstruction using multiple moving cameras in the CARLA driving simulator" has been accepted for publication in the 2022 IEEE International Conference on Big Data.
Achievements
My third Ph.D. research paper titled "Real-time citywide reconstruction of traffic flow from moving cameras on lightweight edge devices" has been accepted for publication in the ISPRS Journal of Photogrammetry and Remote Sensing (IF: 11.774).
Achievements
Awarded 500,000 JPY research fund to work on improving global satellite-based wind models using super-resolution GANs as part of the "Self-directed and Integrated project research" from JST.
Achievements
My second Ph.D. paper on "Citywide reconstruction of traffic flow using the vehicle-mounted moving camera in the CARLA driving simulator" has been accepted for publication at the IEEE ITSC'22 conference, Macau, China.
News
Our accepted proposal for the Big Data Cup challenge at IEEE International Conference on Big Data 2022 is live now. Participate to win cash prizes and complimentary registration to attend the conference in Osaka this December.

Education

IIT Kanpur, India

Degree:
Bachleor's of Technology
(2014-2018)
CGPA:
8.80/10.00
Awards:
Academic Excellence Award (2017),
The Proficiency Medal for the best undergraduate project (2018)

The University of Tokyo, Japan

Degree:
Master of Science
Graduate School of Engineering (2018-2020)
CGPA:
3.87/4.00
Awards:
The Furuichi Kimitake Prize for outstanding Master's thesis (2020)


The University of Tokyo, Japan

Degree:
Doctor of Philosophy
Graduate School of Engineering(2020-2023)
CGPA:
4.00/4.00
Awards:
SpringGX fellowship grant from JST (2021),
Visiting student at Senseable City Lab, MIT



Achievements

Massachusetts Institute of Technology (MIT), USA November 2022

Visiting Ph.D. student

Among top students Internationally to be selected for the visiting student position at the Senseable City Lab, MIT

Japan Science and Technology (JST), Japan September 2021

SpringGX fellowship

Recipient of the SpringGX fellowship provided by JST with a living allowance and research grant of over 25,000 USD per year

The University of Tokyo, Japan September 2020

Furuichi Kimitake Prize

Awarded the Furuichi Kimitake Prize for outstanding Master's thesis among top 4 graduating students (2020) in the department, The University of Tokyo, Tokyo, Japan

ISPRS Geospatial Week 2019, Enschede, The Netherlands Jun 2019

Best Paper Award

Received the Best Paper Award in Laser Scanning workshop during ISPRS Geospatial Week 2019

View more

Internships

Cambridge, MA, USA January 2023 - Present

Senseable City Lab, MIT

Computer Vision Researcher for urban environment monitoring

Los Angeles, CA, USA February 2019 - March 2019

NASA Jet Propulsion Laboratory

Deep Learning Engineer for forecasting precipitation

Heidelberg, Germany June 2018 - August 2018

3DGeo, Heidelberg University

Deep Learning Researcher for 3D pointcloud data

Tokyo, Japan May 2017 - July 2017

Sekimoto Laboratory, The University of Tokyo

Deep Learning Researcher for GPS data

Work experience

Tokyo, Japan 2021 - Present

UrbanX Technologies, Inc.

R&D, Computer Vision Engineer

Tokyo, Japan 2019 - 2022

Ando Hazama Corporation, Japan

DX Strategy Department, Deep Learning Engineer

*All work experiences are part-time only.

Publications

ISPRS Journal of Photogrammetry and Remote Sensing
Real-time citywide reconstruction of traffic flow from moving cameras on lightweight edge devices
Using the TensoRT framework, we optimized the object detection model for inference on lightweight GPU edge devices. Additionally, we develop a client-server framework for estimating traffic flow in real-time on GPU edge devices such as Jetson Nano with comparable accuracy.
READ PAPER
IEEE International Conference on Intelligent Transportation Systems (IEEE ITSC 2022)
Citywide reconstruction of traffic flow using the vehicle-mounted moving camera in the CARLA driving simulator
This research extends our methodology to estimate citywide traffic flow inside a driving simulator. This research helps verify the efficacy of proposed algorithms in the real world under a wide range of environment and traffic conditions.
WATCH PRESENTATION
IEEE International Conference on Big Data 2021
Citywide reconstruction of cross-sectional traffic flow from moving camera videos
This is my first Ph.D. paper, where I developed a new method to estimate traffic flow from moving camera videos such as dashcams and smartphones mounted on cars.
READ PAPER
Public Library of Science (PLoS) One
Convcast: An embedded convolutional LSTM based architecture for precipitation nowcasting using satellite data
As part of my internship at NASA Jet Propulsion Laboratory, I developed a framework called "Convcast," which uses convolutional LSTM networks for short-term precipitation forecasting.
READ PAPER
ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Feature relevance analysis for 3D point cloud classification using deep learning
In the summer of 2018, I interned at Heidelberg University, Germany, and published this paper on identifying relevant features for 3D point cloud classification using deep learning. I won the "Best Paper Award" at ISPRS Geospatial Week 2019 in Enschede, The Netherlands.
READ PAPER

Guest lectures

The University of Tokyo, Japan September 2018 - December 2018

Teaching Assistant for the Deep Learning course

Worked as a Teaching Assistant in "Deep Learning and Related Methods for Large Dataset Information Processing" course taught by Michal Fabinger

IIT Kanpur, India October 2020

Guest Lecture on Deep Learning for Computer Vision applications

A two-day webinar on Deep Learning for Computer Vision applications was delivered remotely at IIT Kanpur, India, with more than 40 participants

Heidelberg University, Germany August 2018

Teaching Assistant for the 3D Geo course

Worked as a tutor for helping students with 3D point cloud processing software for the 3DGeo course taught by Prof. Bernhard Höfle

The University of Tokyo, Japan April 2021 & April 2022

Teaching Assistant for the Geographical Information System (GIS) course

Tutor of the GIS course taught by Professor Yoshihide Sekimoto, responsible for QGIS software handling for map creation, attribute and spatial analysis

© 2023 Ashutosh Kumar

Made with in Tokyo