CodePudding user response:
Small white from the start, check the network to the format, to convert your image to network image format (not only the size, some still can ask converted to double type, etc.), and then training, preliminary Suggestions according to the size of the online, because the general network and form a complete set of training library or have been trained in advance,CodePudding user response:
Data volume is not big, can only use the training model for fine-tuning,