1results about How to "Accurate reference data" patented technology

The present application relates to a kind of methods for predicting porous medium structure using machine learning, comprising: obtaining the image of porous medium prepared in different working conditions and corresponding parameter information;According to phase state characteristics, image is processed, and image dataset is obtained;From image dataset, optionally two images of different working conditions are selected, and with corresponding parameter information, a training sample is formed;Using several training samples, generative adversarial network model is trained, which uses unsupervised image-to-image conversion algorithm for training, for each input training sample, the reconstruction image of original image and predicted image are output, the model is verified by the output predicted image, and the prediction model is obtained after training;The image of porous medium prepared in known working condition and corresponding parameter information are input into prediction model together with the preset parameter information of target object, and the predicted image of target object is obtained, so that the structure prediction of target object can be efficiently and accurately obtained by the present application.