Image processing model training method and device and storage medium

Abstract

The invention provides an image processing model training method and device and a storage medium, an image processing model comprises a backbone network, a regional candidate network and a detection network, and the method comprises the steps: carrying out the feature extraction of a sample image containing a target object through the backbone network, and obtaining a feature map of the sample image; performing region selection on the feature map through a region candidate network, and determining a candidate region; performing target object detection on the candidate area through a detectionnetwork to obtain category parameters and position parameters of a target object, the category parameters including classification results corresponding to the target object, and the position parameters including a bounding box, a segmentation mask and a key point mask corresponding to the target object; determining a value of a target loss function of the image processing model based on the category parameter and the position parameter of the target object; and updating the model parameters based on the value of the target loss function. According to the invention, the target object in the image can be accurately positioned, and the target detection precision is improved.

Description

technical field [0001] The present invention relates to the technical field of machine learning, in particular to a training method, device and storage medium for an image processing model. Background technique [0002] Machine learning (ML, machine Learning) is a branch of artificial intelligence, usually including artificial neural network, belief network, reinforcement learning, transfer learning and inductive learning and other technologies. The purpose of machine learning is to let the machine learn based on prior knowledge, So as to have the logical ability of classification and judgment. The machine learning model represented by neural network continues to develop and is gradually applied to target detection in image processing. [0003] In the related art, the training of the neural network model used to detect the target object in the image is only based on the bounding box information corresponding to the target object in the image, so that the accuracy of the tra...

AI Technical Summary

Problems solved by technology

[0003] In related technologies, the training of the neural network model used to detect the target object in the image is only based on the bounding box information corresponding to the target object in the image, so that the accuracy of the trained image processing model for target detection is low.

Images