Team Status Report for 4/12/2025
This week we made great progress on our project. We successfully enable our system to read images from a video and have an entirely functional computer vision system to detect, extract and recognize the license plates inside those images. We also used new cameras that successfully captures images that are readable for our computer vision software.
James’s Status Report for 4/12
I have been working on getting the CV pipeline working for videos, which was successfully accomplished. The only following 2 steps are making sure it works with the Jetson and getting the GPU running for it, then getting it working with the camera stream. At the moment, it’s working with videos taken from the internet and pushed through the pipeline, at about 17 fps.
Xinglan’s Status Report for 4/12
This week I was working on accelerating our computer vision pipeline. We used to have a working pipeline that was really slow, Which I managed to accelerate with a better OCR. Then I worked on combing it with the video reading and plate detection module built by Josh, as well as integrating a faster plate detection system into the Jetson environment.
Joshua’s Status Report for 4/12
These past two weeks I have been focusing on getting the display looking good, and integrating it in with the rest of the pipeline. Now users are able to easily enter license plate information, and it should be easy to display for the user as well.
I also spent a large amount of time debugging the arducam issue, but to no avail. I’m convinced at this point it is a hardware issue, so I have ordered new camera ribbon strips to see if it fixes the issue. In the meantime we’ll be using a normal USB web camera to test the pipeline.
Team Status Report for 3/29/2025
Xinglan’s Status Report for 3/29
This week, I have tried various ways of implementing our computer vision pipeline on Jetson. A lot of the libraries and modules involved were trickier to install on Jetson, such as Pytorch, which will not work will with CUDA if I only install a normal version as shown on their original site. But eventually I am able to setup the entire CV pipeline with license plate location and extraction from image, then character recognition. The performance of our models though still needs improvement.
GANTT Chart
James’s Status Report for 3/29
This week, I successfully implemented the GPS reading system into the jetson orin nano, however integration still must occur later on. I helped Xinglan with torchvision debugging, and I restructured the GANTT chart to help us meet our goals better for the end of the semester. Finally, I worked with Xinglan on cropping license plate images as well, in order to help the OCR detection.
Team Status Report for 3/22/2025
Our team worked with Professor Tamal to figure out the flashing Jetson problem. We have now sufficiently flashed the Jetson with the correct Jetpack using an old linux-based computer. This will allow us to start running our CV pipeline directly on the device, as well as connect our arducam with the Jetson to interact with the display. The next easiest form of testing for our device will involve finding the best lense for our use case (given a base accuracy of 90%).
Xinglan’s Status Report for 3/22
This week I worked on setting up all the CV toolkit on Jetson. The implementation of the license detection and recognition network required a lot of peripheral packages, and a lot of documentations are out of date. I fixed many of these problems and managed to download the main CV application. However, the runtime environment on the Jetson still requires a few additional fixes.
Josh’s Status Report for 3/22
This week, I helped James try to troubleshoot the Jetson startup. I had no luck on my end, which honestly just helped eliminate dead ends from our solution path. I was not as helpful this week because I was out of PGH from Wednesday-Sunday, in addition to the ethics brief. Hoping to get some hardware configured for next week.
James’s Status Report for 3/22
This week, I met with a professor to finally get the flash working for the Jetson Orin Nano, and finally got it to work. I began to work on getting the jetson SSH’d onto our computers so that we can work with a lot less cords and still accomplish tasks. We can now move onto testing out license plate networks and identifying characters.
Team Status Report for 3/15/2025
We started work on setting up the CV, as well as continuing to debug our Jetson setup. We’ve come up with some additional plans of action, including startup, different versions, and even working with a different Jetson. We will pursue these as our timeline gets closer and closer. In addition, the right components have arrived and are board ready for development and testing.
James’ Status Report for 3/15
This week, I did even more troubleshooting on the Jetson Orin Nano setup, as it is holding up the entire project, and we may need to acquire a new device. I completed the ethics assignment, and did more computer vision pipeline research. At this point, we may order a new Jetson Orin Nano if Monday’s testing doesn’t provide any changes in the effectiveness of the device.
Xinglan’s Status Report for 3/15
I was working on getting our CV algorithm running on my personal computer before implementing in on Jetson. The related toolkit and packages were not as straightforward to install on WSL. However, since we have some problem setting up our Jetson and my personal computer has a CUDA compatible Nvidia GPU, I will use my computer this way to test our algorithm.
Joshua’s Status Report for 3/15
Finally got some working components in! I have been working with the display and display encoder, as well as our new arducam camera module. The display works just fine, and is actually working as it should be in displaying for the Jetson. The camera connects just fine, although we will need some more development via the Jetson in order to get some testing/development in there.
Team Status Report for 3/08/2025
This week was focused on getting finished with our CV pipeline so we can begin testing it, as well as finalizing all designs and purchases for future use. By encountering issues and bugs this week, we can continue to work on and develop on these problems next week.
Xinglan’s Status Report for 3/08
I have been trying to verify our proposed LPRnet that we intended to work with YOLO. The related toolkit and packages involved have been more complicated than expected. The setting up process took hours to navigate through and debug. I also wrote much of the design report and looked at much related research of our project in the process.
James’s Status Report for 3/08
I have been working towards completing the Jetson Orin Nano setup and getting a simple MVP working, so that serious alterations, tweaks, and camera setups can take place. This has been proven to be a tedious process, as there have been a lot of setup problems and issues in the meantime. I have worked on multiple parts of the design report as well in order to complete it by the due date and understand the place that our group stands in terms of previous projects and the actual pipeline of our project. We will now be using this as the foundation moving forward in the project.
Josh’s Status Report for 3/08
I have been working towards getting our final states of hardware so that the project is essentially ready to put together once the Jetson is working. I have done research on the camera and found appropriate lenses we can use for testing. I have also done research on a display solution that will allow us to utilize touchscreen to input license plates, and to interact better with our device. At a measly 5″, it is small enough to work with our device. In addition, the encoder allows us to seamlessly connect the device with our Jetson via HDMI.
Team Status Report for 2/22/2025
This week was focused on the design presentation, developing our MVP, starting our Written design report, ordering parts, and considering metrics for later testing.
James’s Status Report for 2/22
This week, we worked on the design presentation, and I began our design report as well. I started working on research for the Jetson Orin Nano implementation, as in how to get it up and running, and connected to a laptop. I also prepared the presentation speaking, as I was the one that was presenting this week.
Joshua’s Status Report for 2/22
This week I helped finish the design presentation, as well as started to look at a lot of the hardware for our design. This week I worked on finding some of our components for the MVP, looking at lenses and configurations for our camera, and filling out the ordering form via canvas.
Xinglan’s Status Report for 2/22
This week we finished the design presentation and got feedback about our design and component choices. We are experimenting with our arriving components to address these problems. We are considering the resolution of our cameras and redecided our detection ranges and recognition metrics. I started working with our embedded vision board Jetson, and trying pretrained algorithms.
Team Status Report for 2/15/2025
Our team worked on the design presentation, and focused a lot of our efforts into product research and orders, especially for the necessary parts to reach MVP.
Xinglan’s Status Report for 2/15
I am working on the design presentation this week. Our team is researching many design choices, including software solutions for our computer vision implementation and peripheral hardware of the system. Thus, I have to keep our design review updated with our choices of solutions and explain our approach of integrating them into our MVP. We have to explain the specific of our workflow with our modified approach, such as locating the license plate and processing the GPS locations.
James’ Status Report for 2/15
Most of my time was spent researching cameras, CV algorithms, and IMU sensors. I found 3 separate approaches to the problem of motion blur, the first being optical image stabilization (OIS), the second being software-based, and the third being IMU-based. I quickly ruled out software-based due to processing constraints and desiring real-time computation. From there, I focused on finding products for these 2 separate implementations. The OIS camera is very all-in-one, with great image quality and FOV, an IR-cut filter, and HDR support all while being able to accomplish 30FPS. The IMU approach is cheaper, but I’d still need to get a camera that checks off all of the boxes outside of OIS, which would basically even out the costs.
Outside of product research, I researched CV algorithms, Github repo’s, and implementations, and created a CV pipeline to be refined later, but it gives a good general structure to begin our MVP with.
Finally, I worked on the Design Presentation, and made sure the Solution Approach included the research that I did on cameras and CV algorithms.
Joshua’s Status Report for 2/15
Team Status Report for 2/8/2025
Our team is beginning to make considerations that will be important to our design preview/presentation. This includes finalizations of components/hardware used, as well as the background research to show that our desired software/hardware use can work together to create our MVP.
Xinglan’s Status Report for 2/8
I looked into the requirements of our project and researched the software and hardware components of it. Then I started writing the design presentation based on our current plan of the project, including requirements and use case. I also worked on getting more detailed description of our system since we have figured out some of our specific components for the job.
James’ Status Report for 2/8
I did research into peripherals such as the AMD vision board, GPS systems, and IMU devices. I looked into stabilization of motion blur for CV algorithms, and how to go about doing that.
I also analyzed other groups’ presentations and took feedback on our own, in order to remap the GANTT chart and reach MVP sooner. Then we can build from there for a greater project in scaling.
Joshua’s Status Report for 2/8
No significant deliverables for this week. Most work has been doing background research on deploying the Jetsin Orin Nano, as well as beginning to look at the hardware we may use for our peripherals.
In terms of hardware there’s several components that we may need to install. We have a GPS package, including dead reckoning (IMU), a camera, a display, and a potential ESP. The Jetsin and the camera have been secured via ECE Inventory. We’re working on securing other components.
I will also work on creating a pcb board to interact with (ideally) the gpio header pins on the Jetson to simplify deployment. This will make it easier to test and attach our components, and hopefully simplify our final MVP. It may also be interesting for me to work on some physical components (like laser cut wood) to contain our peripherals and JETSIN system.