February 1, 2021 Update

Some 3D parts need to be reprinted to fit the camera module better. Specifically, the camera housing has to be enlarged a little and the wire hole position adjusted slightly. The changes and printing should be done within two weeks.

For power, our initial power solution has some flaws making it impossible to provide enough power, causing the Jetson to brownout. In order to continue development at home, we've solved this by using an AC adapter. However, the issue still remains when we try to power the nano with a car's cigarette adapter. We are working on alternative solutions to power the computer with the car.

We've successfully programmed the nano to take in three camera input streams. However, we are currently limited to 22 FPS. This limit does not change even as we change the resolution of the cameras - indicating some sort of bottleneck elsewhere. We are currently attempting to make use of Python's multiprocessing library to improve performance.

Along with the camera streams, we have also added the ability to switch between our blindspot detection mode and our rear-view mode. This is simply us switching between the appropriate camera streams in software.

Once the camera stream's FPS have improved, we will be able to connect the UI to the model's output, completing the software portion of the project.

January 28, 2021 Update

Quick update meeting on progress of getting multiple camera stream intakes working + UI work.

Multi-threading the cameras is proving a little difficult as NVIDIA has their own custom Python-wrapped C++ code objects that throw errors when trying to pass objects between multiple threads. Working on a work-around/solution now.

Mock UIs have begun development and will go through review cycle soon.

January 21, 2021 Update

Met with our prof advisors today as well as our normal meeting.

Not too much with the professors, just more of an update as to what's expected as deliverables as well as briefing them on our plan and current progress, as well as what's to come.

On the group's end, Jonathan has finished setting up the Jetson Nano to handle machine learning tasks. The mirror mounts have also been prepped by Tian to install the camera modules inside them. Eric is currently working with the Jetson and cameras to set up the multiple streaming inputs we will need. Unfortunately, due to the lockdown, only one group member can have a specific part at a time, making it difficult to test the cameras with the mounts while they are being setup with the Jetson. Lastly, for the UI component, Justin is working on some rough preliminary designs in OpenGL.

We've also encountered some potential problems with our current power solution. Our purchased adapters don't work as expected and as a result, don't provide enough power for the Jetson. We're working on various potential solutions to remedy this now.