Hi,
I’m training a YOLO model in Roboflow for enemy detection, and I want the model to not detect teammates/friends at all (avoid false positives on friends)
Do you have a class in your dataset for friends? if so make sure that the data set follows a clear rule of what classifies a friend vs. an enemy. Add more examples especially of situations where differences are hard to see.
Hi Mohammad!
This seems like a really cool project! I think one key note here is that there does need to be two classes, friendlies, and enemies. The other thing I would ask yourself is if you looked at the data, would you be able to tell the difference between a friendly and an enemy, or in other words are their visual differences between the two (maybe an outline or different colored text above their heads). Those sort of differentiators are very important when we are teaching a new computer vision model what to ‘look’ for and ultimately end up with a good result. Hope this helps you out!