I'm really new to object detection with Yolov3. Let's say I have 10 classes and the amount of data is approximately the same. Do I achieve better average precision when I use 10 Yolo models and train them always with the data of just one class or will there be a higher average precision if I only have one model I train it with the data of all classes?
For me it's hard to know because on the one side a single-class model can focus on detecting just the features of one class but on the other side if I use only one model, the model learnes better how to distinguish between the classes, right?
What's your opinion? And does it changes in the case of imbalanced number of images per class? If I have 60000 images of class 1, but only 5000 of the other 9 classes, a model which tries to detect all objects will shift to the class 1 in the training right? So are single-class models better in this case?
Thanks for your opinion. Cheers!