Adam Beckermann
Adam Beckermann
Language of Study: Spanish
Subject Major: Computer Engineering
Country of Study: Spain
Host University: Universidad Carlos III de Madrid
Title of Work:
Autonomous Vehicles and their Perception Systems
Abstract:
My internship occurred at Universidad Carlos III de Madrid, where I worked in an autonomous vehicle laboratory. My internship revolved around the perception systems within an autonomous vehicle, which is the system that detects vehicles, objects, people, and any other thing that comes close to the vehicle. I was tasked with learning a new program called YOLO (You Only Look Once) v5 as it was an extremely new object detection program that needed to be tested to see if it could be used precisely in perception systems. My main goal was to test how precise YOLOv5 was in comparison to previous versions of YOLO. At the end of my internship, the information I had gathered was utilized in a presentation for a company in Madrid named Sacyr. The future of autonomous vehicles, which I hope to pursue, is extremely important due to their environmentally friendly nature and high degree of safety.
Title of Work (IGP Language):
Vehículos Autónomos y sus Sistemas de Percepción
Abstract (IGP Language):
Mi pasantía fue en la Universidad Carlos III de Madrid, donde trabajé en un laboratorio de vehículos autónomos. Un vehículo autónomo es un vehículo con la capacidad de operar sin ninguna intervención humana. Mi pasantía fue trabajar con sistemas de percepción para vehículos autónomos, que es el sistema en un vehículo autónomo que detecta cosas alrededor del vehículo. Aprendí un nuevo programa llamado YOLO v5 ya que era un programa extremadamente nuevo y comprobé si se podía usar. El objetivo de mi trabajo era ver si YOLO v5 debería implementarse en los sistemas de percepciones de los vehículos autónomos y si era más preciso que las versiones anteriores. Una empresa llamada Sacyr utiliza mi información para continuar su desarrollo de vehículos autónomos. El futuro de los vehículos autónomos es extremadamente importante debido a que son ecológicos y reducen la tasa de accidentes y es algo que quiero perseguir en el futuro.
Elevator Pitch Transcript:
Hello everyone. My name is Adam Beckermann. I'm here today to present my elevator pitch for IGP Summit Fall 2022. The title of my work is Autonomous Vehicles and their Perception Systems. A little bit about me. My subject major is computer engineering and my language of study is Spanish. I studied abroad in Madrid, Spain, a la Universidad Carlos Tercero de Madrid, and I worked at UC3M within their Intelligence Systems laboratory for my internship. Quick outline of my experience. I started with creating a database within Scalable of images with bounding boxes around them that showed where cars were. These images were added to a larger data set, about eight to ten thousand to be utilized later. I was then tasked with learning a program called YOLOv5, which stands for you only look once version five. Once I learned how to utilize it, I utilized YOLOv5 with our large eight to ten thousand image data set to analyze the data and utilize these bounding boxes to create videos and other things of that nature to see how precise YOLOv5 is and see if we can use it within perception systems. Then, I compared YOLOv5 with previous versions, YOLOv4 and YOLOv3 primarily, to see which one was the best. YOLOv5 ended up being the most precise, but they didn't have the support that the other two have. So in the upcoming years, YOLOv5 will be very prominent. Once all this was found out, I was tasked with improving mAP, which is mean average precision as much as possible within YOLOv5. If you look at those images on the bottom right, you can see on the very far right column, a very poor mAP for the first data set. And then once the data set was changed, some various things were changed, it was much improved in the second picture.Then all my data was submitted to be utilized within a presentation for a company named Sacyr to improve and continue to develop. So what was so interesting about my internship? I thought being the first person to teach a wide range of people about a new program within object detection was really something cool to me and something I was really proud of. Along with that, being on something such as revolutionary as autonomous vehicles and perception systems is something that's extremely interesting to me. What I think would be interesting to other people is these videos we're able to produce utilizing my data. So what you're seeing here is what an object detection or what a perception system would see. They see the cars, they see the motorcycles and other things of that nature. And this just shows how precise the perception system would be at dealing with these things if they're close to it. Obviously, within a perception system, there will be cameras all around the car, so it wouldn't be this one that's very zoomed out, it would be one that's much closer. So you can see even though it's pretty zoomed out, it's pretty accurate. And then the future of my work is one that's extremely important and something that I hope to pursue. Autonomous vehicles are very environmentally friendly and much more safe. So, I hope to be a part of the cutting edge when it comes to autonomous vehicles and I hope to see you at my summit presentation. Thank you.