Rock classification constraint inheritance loss method for rock recognition deep learning model

The invention discloses a rock classification constraint inheritance loss method for a rock recognition deep learning model. The method comprises the steps as follows: a rock sample data set D is constructed according to multi-level labels of rock samples, a deep neural network is constructed, rock sample images to be trained are input into the deep neural network, and the network output predictsthe last-level label of each level of a classification tree; the prediction probability Pi of each level of the classification tree is calculated; step-by-step calculation is performed on the prediction probability Pi of each level of the classification tree according to the corresponding level label to obtain cross entropy loss of each level; the cross entropy loss of each level is added to obtain rock classification constraint inheritance loss of the input rock sample image; and whether the rock classification constraint inheritance loss meets a convergence condition or not is judged. The method has the advantages t