I have trained a model based on YOLOv5 on a my custom dataset which has nine classes (for example class0 ~ class8).
And then, I tried to detect it with best.pt trained with only my custom data.
I wanted it to predict 89 classes, including the existing 80 pretrained classes, but it only predicts the 9 classes I trained.
I suspected overfitting and lowered the epoch, but the prediction probability of only 9 classes dropped and the existing 80 classes were not detected.
There are 89 classes in the YAML file, and the data is only my data (with 9 classes). Is there something I need to set up differently, how can I solve this problem?