I have a question about training object detection models especially using the YOLOv7 algorithm.
I use a Soccer Players Dataset from the Roboflow. The dataset is very unbalanced.
I trained the model using basic hyperparameters and it detects all objects on images but the predicted label is invalid. The model predicts everything as a player (but for example one person on an image is a referee).
How can I improve label prediction accuracy?