Image processing model generation method, image processing method and device, and electronic device

An image processing model generation method, comprising:

• inputting at least one training sample lesion image of at least one known disease type into an initial image processing model, wherein the initial image processing model includes a classification layer and a marking layer, and each training sample lesion image includes coordinates of an initial center point of a lesion region in the training sample lesion image, an initial length of the lesion region, and an initial width of the lesion region;
• calling the classification layer to perform classification processing on the training sample lesion image, so as to obtain a classification probability of a known disease type corresponding to the training sample lesion image;
• calling the marking layer to process the training sample lesion image, so as to obtain coordinates of a predicted center point of the lesion, a predicted length of the lesion region, and a predicted width of the lesion region;
• obtaining a loss value of the at least one training sample lesion image in the initial image processing model, according to the classification probability corresponding to the training sample lesion image, the coordinates of the initial center point, the initial length, the initial width, the coordinates of the predicted center point, the predicted length, and the predicted width;
• and
• determining whether the loss value is within a preset range;
• if the loss value is not within the preset range, updating parameters of the initial image processing model, wherein an image processing model with updated parameters is used as an initial image processing model in next training;
• and
• above steps are repeated until the loss value is within the preset range, and an image processing model in last training is used as a trained target image processing model.