Most of my time this week was spent preparing for the proposal presentation, with some research in the CV system, as well as looking into parts we may want to order. We received some feedback from the presentation about the YOLOv4 tiny model that we planned to use, with a suggestion that we could try a more recent YOLO version, or even a different model entirely that is more accurate (ie. Faster R-CNN). Since YOLO is rather easy to set up, I am currently working on setting up YOLOv7 to try training on a custom dataset. Our order for a Jetson from the 18500 inventory has been approved, so hopefully I can try deploying YOLO on it soon. I have found multiple datasets of recyclable materials (ie. https://www.kaggle.com/datasets/alistairking/recyclable-and-household-waste-classification), so we have plenty of training data available.
I also looked into serial communication between the Arduino and Jetson. Both devices have serial communication pins that can be used to talk with each other. It is also possible to connect the two with a micro usb cable, which will likely be our approach. Communication between the two devices is an important part of meeting our project’s use case, so it’s good to see that this is possible and has been done before.