Development and Research of Methods for Detecting Distorted Data Using a Multimodal Approach

The paper considers the problem of detecting distorted data using a multimodal approach. Various methods of classification are being investigated. In the course of experimental studies, it was found that the best results are demonstrated by the method based on the pre-trained multimodal RUDOLPH model. The highest value of the F1 metric is 0.841.