Method for image classification, computer device, and storage medium

A computer device, comprising a memory and a processor, the memory storing computer-readable instructions which, when executed by the processor, causing the processor to perform the steps of:

• acquiring a to-be-classified image and inputting the to-be-classified image to a trained image classification model, the trained image classification model comprising a localization segmentation sub-network, an alignment sub-network, and a classification sub-network, the alignment sub-network being formulized as a valve linkage function, the image classification model being obtained by training by adjusting parameters of the localization segmentation sub-network and parameters of the classification sub-network by the valve linkage function, and in a forward-propagation phase of the training, an output of the valve linkage function being an aligned image;
• in a backward-propagation phase of the training, the output of the valve linkage function being a function with respect to a localization region and a segmentation region outputted by the localization segmentation sub-network;
• subjecting the to-be-classified image through the localization segmentation sub-network for locating and segmenting a target object of the to-be-classified image to obtain a segmented image comprising a localization region and a segmentation region;
• subjecting the segmented image through the alignment sub-network, the alignment sub-network aligning the target object to obtain an aligned image;
• and
• subjecting the aligned image through the classification sub-network for fine-grain classification to obtain a class corresponding to the to-be-classified image.

One or more non-transitory storage medium storing computer-readable instructions which, when executed by one or more processors, causing one or more processors to perform the steps of:

• acquiring a to-be-classified image and inputting the to-be-classified image to a trained image classification model, the trained image classification model comprising a localization segmentation sub-network, an alignment sub-network, and a classification sub-network, the alignment sub-network being formulized as a valve linkage function, the image classification model being obtained by training by adjusting parameters of the localization segmentation sub-network and parameters of the classification sub-network by the valve linkage function, and in a forward-propagation phase of the training, an output of the valve linkage function being an aligned image;
• in a backward-propagation phase of the training, the output of the valve linkage function being a function with respect to a localization region and a segmentation region outputted by the localization segmentation sub-network;
• subjecting the to-be-classified image through the localization segmentation sub-network for locating and segmenting a target object of the to-be-classified image to obtain a segmented image comprising a localization region and a segmentation region;
• subjecting the segmented image through the alignment sub-network, the alignment sub-network aligning the target object to obtain an aligned image;
• and
• subjecting the aligned image through the classification sub-network for fine-grain classification to obtain a class corresponding to the to-be-classified image.