This past week, I spent most of my time coalescing the datasets of the trash items we wish to detect into respective classes and also thinking/researching ways to further tune the YOLOv7 and YOLOv7-tiny models. As I’m worried about wasting compute resources, I want to be careful about how many training runs I’m doing, and thus am waiting on training advice before making any major changes. In the meanwhile, while I’m blocked, I’ve been putting together code for the overall ML pipeline. Additionally, we’re still waiting for quite a few components to come in (including an SD card reader to flash our microSD card), so we’re stuck on progress in terms of working with the Jetson Nano Orin. We also had design review presentations in class and began working on our design review report.
Since I can’t get onto the Jetson Nano Orin quite yet and I’m blocked on tuning, I’m behind schedule. However, I should still be on track to finish the ML pipeline and tuning our model by the end of Spring Break, as per our gantt chart, since I’ve more or less swapped the order of my tasks. As a team, we will also complete the design report next week.
0 Comments