Evaluation method, system and equipment for deep counterfeiting detection model

Abstract

The invention belongs to the field of image processing, and discloses an evaluation method, a system and equipment for a deep counterfeiting detection model. Comprising the following steps: according to a training method of a to-be-evaluated deep counterfeiting detection model, respectively training the to-be-evaluated deep counterfeiting detection model through preset deep counterfeiting data sets of various types to obtain each trained deep counterfeiting detection model; testing each trained deep forgery detection model through a preset diversified difficult sample set; obtaining an accuracy index value of the trained deep forgery detection model under the same distribution data, and a generalization index value, a robustness index value and a practicability index value under the different distribution data; and carrying out weighted superposition according to a preset weight to obtain an evaluation result of the to-be-evaluated deep counterfeiting detection model. The accurate, fair and comprehensive evaluation method is established, and the obtained evaluation result is more in line with the actual situation of the deep counterfeit detection model.

Description

technical field [0001] The invention belongs to the field of image processing, and relates to an evaluation method, system and equipment for a deep forgery detection model. Background technique [0002] In recent years, the continuous development and innovation of artificial intelligence technology represented by deep learning algorithms has led to continuous breakthroughs in the solutions to many tasks in the field of computer vision. Its successful application has brought many conveniences to life and social production, such as intelligent video surveillance scenarios, Autonomous driving scenarios, smart medical scenarios, etc. However, the abuse of such technologies may pose a huge challenge to the protection of personal privacy. The recently proposed deep forgery technology (DeepFake) based on deep learning misleads people to believe the false words and deeds in the video by tampering or replacing the face information of the original video. New threats to invasion of pr...

AI Technical Summary

Problems solved by technology

However, because it only evaluates the methods that appear in the competition, and does not set any restrictions on the training process of the participating methods, it lacks strict and fair evaluation of the existing mainstream detection methods

[0004] In addition to this, the current evaluation of deepfake detection models is unfair and inadequate for the following reasons, resulting in inaccurate results

First, many evaluations of deepfake detection models use models trained on different training deepfake datasets for evaluation. For example, some methods directly apply publicly available trained models during evaluation, instead of using the same Re-implementing these models and evaluating them on the training data of the deepfakes, this inconsistent evaluation effort on training deepfake datasets can lead to unfair and incorrect comparisons between methods

Secondly, since most deepfake detection models are trained and evaluated on the same distribution deepfake datasets generated by only finite forgery generation methods, there are problems of overfitting and poor transferability, which lead to most The performance of the detection model with seemingly excellent performance is greatly reduced when it is actually applied in the real scene.

Images