Following this tutorial, I trained YOLOv8 on a custom dataset to detect holes in the ground. I obtained acceptable results considering how small my training set is (about 50 pictures).
When preparing my dataset, I also labelled objects such as persons or trucks:
.
Person or truck detection did not work as well, probably because I only included a handful of relevant images in my training set. But from what I read, YOLOv8 should already be "pre-trained" for these two classes.
So:
- Should I re-train YOLOv8 to identify pre-trained classes?
- If not, how can I ask YOLOv8 to identify specific classes even though they were not labelled in the training set?
