SMART INTERSECTION

2020 Summer Internship Program

Rutgers University
Wireless Information Network Laboratory
OBJECTIVE

The goal of this project is to create a method for estimating vehicle count/traffic flow statistics for a given New York City intersection. YOLOv3 object detection running as an NVIDIA DeepStream application is used to count vehicles as they enter and exit intersections from various directions. Traffic flow is analyzed for comparison with flows from different days and time periods. Inference output is published to an AMQP message broker and can be recombined with the original video stream on separate machines via OpenCV.

WEEKLY PROGRESS

Aeriel View of New York
WEEK 1

June 1 - June 5

Bustling New York
WEEK 4

June 22 - June 26

Skyline New York
WEEK 7

July 13 - July 17

Sunset Over New York City
WEEK 2

June 8 - June 12

New York City
WEEK 5

June 29 - July 3

Skyscrapers Above Times Square at Dawn
WEEK 8

July 20 - July 24

New York City
WEEK 3

June 15 - June 19

New York City
WEEK 6

July 6 - July 10

Brooklyn Bridge
WEEK 9

July 27 - July 31

 

PUBLIC PRESENTATIONS

Public Presentations are presentations compiled and presented to general audiences who may not be familiar with our technologies and their technical details.

Sunset Over Manhattan
Open House Presentation

 

Check out our project poster and watch a video recording of our presentation
on the WINLAB 2020 Summer Internship Open House page!

INTERN CLASS

PRESENTATIONS

Intern Class Presentations are presentations compiled and presented to the entire intern class to confer knowledge gained about tools and software used by multiple internship projects.

Road and Bridge Network
DeepStream and YOLOv3
Overview + Demonstration

 
Urban Traffic
Project Review

 

TEAM PRESENTATIONS

Team Presentations are presentations compiled and presented to advisors during internal team meetings to deliver project updates and share new discoveries.

Security Camera
Introduction to DeepStream, YOLOv3, and 3D Point Cloud Object Detection

 
CCTV Camera
Investigation of DeepStream YOLOv3,
YOLOv3 Using Depth Maps,
and Complex-YOLO

 
Video Camera Lens
Investigation of YOLOv4 and Its Use With TensorRT and ONNX Model Conversion

 
Server Installation
Pub/Sub With ZeroMQ and Recombining Inference Output With Input Video Stream

 
Servers
Pub/Sub Pattern In Depth, Syncing Machines via NTP, CZMQ Publishers, and Bounding Box Information with OpenCV
 
Digital Network Cables
Update on DeepStream Publisher, The Subscriber Class Implementation, Update on OpenCV Developments
 
 
 

MEET THE TEAM

Nicholas.png

NICHOLAS MEEGAN

Rutgers University - Class of 2021

Major:

- Computer Engineering

Minor:

- Computer Science

  • LinkedIn
  • GitHub
20200726_142821.jpg

KEVIN ZHANG

Rutgers University - Class of 2022

Major:

- Computer Engineering

Minor:

- Statistics

  • LinkedIn
  • github-logo
zhubryan_photo.jpg

BRYAN ZHU

Rutgers University - Class of 2021

Majors:

- Electrical Engineering

- Computer Science

  • LinkedIn
  • GitHub

ACKNOWLEDGEMENT

We would like to thank the following individuals for advising and guiding us over the course of our WINLAB internship.

Ivan Seskar, Associate Director at WINLAB

Zoran Kostic, Associate Professor at Columbia University

Jennifer Shane, Laboratory Engineer at WINLAB