Fig 2 - available via license: Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International
Content may be subject to copyright.
Attribute annotation correlations of the DeepFake detection databases -The 20 most positive and negative (Pearson) correlations are shown for each of the five databases. Green indicate positive correlations, while red indicates a negative correlation. The highly-correlating attributes should be considered for working with these databases to prevent misinterpretations.
Source publication
In recent years, image and video manipulations with DeepFake have become a severe concern for security and society. Therefore, many detection models and databases have been proposed to detect DeepFake data reliably. However, there is an increased concern that these models and training databases might be biased and thus, cause DeepFake detectors to...
Contexts in source publication
Context 1
... Attribute Correlations: To understand the quality of the labels and potential biases in the attribute space, Figure 2 shows the 20 most positive and negative pairwise attribute correlations. For instance, we notice in Figure 2a that the attributes of Mustache and Goatee are highly correlated with each other. ...
Context 2
... Attribute Correlations: To understand the quality of the labels and potential biases in the attribute space, Figure 2 shows the 20 most positive and negative pairwise attribute correlations. For instance, we notice in Figure 2a that the attributes of Mustache and Goatee are highly correlated with each other. A high correlation also occurs between Heavy Makeup and Wearing Lipstick. ...
