Default yolov4 tiny .cfg doesn't include mosaic augmentation, were any experiments done with it?
Yes. mosaic degrades accuracy for tiny model.
Stronger data augmentation should be used for large models.
Yes. mosaic degrades accuracy for tiny model.
Stronger data augmentation should be used for large models.
What is the difference between "tiny model" and "large model"?
(If it's custom model, )
Is that mean the number of the classes?
or size of the training dataset?
@smallerhand Large model is referring to normal YOLO v4 ( large because-> much more layers & computations than tiny version) The training size dataset doesn't change the computations of the CNN (it may take longer to train tho)
@AlexeyAB Thanks)
Most helpful comment
Yes. mosaic degrades accuracy for tiny model.
Stronger data augmentation should be used for large models.