The Jetson Project

Project Description:

There are many situations in which you may need to surveil an area and watch for certain events to occur. Perhaps you need to know how much traffic travels through a particular road and collect that data to find out how busy that road is on a given day, or maybe you need to watch for corroded pipes at a facility and report their location to a technician so they can be replaced. These situations brought a need for the system we built, which is an object detection application that can detect instances of objects such as people and cars within digital images and videos. In addition to object detection, we’ve also implemented basic analytics to count the number of objects of each type in the video and live stream.

Over the course of 8 months, we faced many challenges. At first, the project was going to be a single application, then it developed to become a full stack web application which resulted in many changes that had to be made, as well as challenges with the transfer of data from one application to another. We utilized modern technologies to build our full-stack web application. We used ReactJS since it’s a well-known and supported web front-end technology. Our back end consists of a Spring Boot application connected to a MongoDB database. It was important to have a non-relational database since the data we’re passing is very dynamic. Finally, machine learning and object detection are best performed using Python as it fully supports all the libraries and functions required by our project.

While we have overcome and came up with solutions for some of the challenges that we faced, there are still objectives that we couldn’t achieve. Our Jetson project was initially planned to work on a drone with a Jetson Nano device attached to run our object detection Python code. However, because of the travel restriction due to the COVID-19 pandemic, we didn’t have access to the device and the ability to work together on the hardware. For that reason, our code is only developed and tested on our BYOD. Secondly, the project infrastructures are running mostly on Google Cloud Platform and would be costly in the future to operate. For now, we have been conservatively using our limited trial credit for this project, which will be expired this December. Thirdly, our detection code still has some major problems with its accuracy; more time is required to work on this problem in the future. Last but not least, we planned to add analytics functionality to the app, but as our project has multiple layers of applications that are written in different technologies, setting and maintaining the connections between them consumed most of our already limited time.

Contact the Team

Funded By