Density network model training method and device, computer equipment and storage medium

Abstract

The invention relates to the technical field of image processing, and provides a density network model training method and device, computer equipment and a storage medium. The method comprises the following steps: determining a model downsampling factor and an image scaling multiple of a density network model to be trained; determining an image scaling parameter according to the model downsamplingfactor and the image scaling multiple; scaling the labeled training sample original image according to the image scaling parameter, and converting the scaled labeled training sample original image into a training sample density image; loading the training sample density image and the unlabeled training sample original image to the density network model, and training the density network model based on the down-sampling factor of the model. The prediction reasoning performance of the density network model can be improved, and the training time can be shortened.

Description

technical field [0001] The present application relates to the technical field of image processing, in particular to a training method, device, computer equipment and storage medium of a density network model. Background technique [0002] With the development of computer technology, there has been a technology for image analysis and processing using deep learning network and other density network models. In specific training, the density network model needs multiple rounds of training. If the image used for training is loaded into the density network model, the density network model involves the process of scaling the image and generating the corresponding density image. , but for images with small and high density, multiple occlusions, diverse backgrounds and a lot of interference, the processing of scaling and generating density images in the density network model will cause the integral value of the density image to change, and it is difficult to guarantee the prediction ...

AI Technical Summary

Problems solved by technology

In specific training, the density network model needs multiple rounds of training. If the image used for training is loaded into the density network model, the density network model involves the process of scaling the image and generating the corresponding density image. , but for images with small and high density, multiple occlusions, diverse backgrounds and a lot of interference, the process of scaling and generating density images in the density network model will cause the integral value of the density image to change, and it is difficult to guarantee the prediction accuracy of the density network model ; Exemplarily, when the semen image under the microscopic field of view is input to the density network model for training, if the density map conversion and training are performed on the semen image in the density network model, it will cause the density network model to predict the sperm in the semen image The statistical prediction accuracy of the quantity is reduced; it can be seen that there is a technical problem of low prediction performance accuracy of the density network model

Images