Methods and apparatus to detect deepfake content
Methods, apparatus, systems and articles of manufacture are disclosed to detect deepfake content. An example apparatus to determine whether input media is authentic includes a classifier to generate a first probability based on a first output of a local binary model manager, a second probability based on a second output of a filter model manager, and a third probability based on a third output of an image quality assessor, a score analyzer to obtain the first, second, and third probabilities from the classifier, and in response to obtaining a first result and a second result, generate a score indicative of whether the input media is authentic based on the first result, the second result, the first probability, the second probability, and the third probability.
1. An apparatus to determine whether input media is authentic, the apparatus comprising:
a classifier to generate a first probability based on a first output of a local binary model manager, a second probability based on a second output of a filter model manager, and a third probability based on a third output of an image quality assessor; and
a score analyzer to:
obtain the first, second, and third probabilities from the classifier; and
in response to obtaining a first result and a second result, generate a score indicative of whether the input media is authentic based on the first result, the second result, the first probability, the second probability, and the third probability.
2. The apparatus of claim 1 , wherein the input media is authentic when the score is less than a score threshold.
3. The apparatus of claim 1 , further including a filter model manager to generate the second output by:
applying a Gabor filter to a frame of the input media; and
extracting signals from the frame of the input media responsive to the Gabor filter being applied, the signals corresponding to changes in texture in the input media.
4. The apparatus of claim 1 , further including a blur detection model manager to generate the second result, the second result representative of a degree of blur in frames of the input media.
5. The apparatus of claim 1 , further including an eye detection model manager to generate the first result, the first result being a binary value indicative of whether a blink occurred in a human face in the input media.
6. The apparatus of claim 5 , wherein the eye detection model manager is to determine that the blink occurred when a ratio of points in an eye satisfies a threshold for four frames of the input media.
7. The apparatus of claim 1 , wherein the first, second, and third probabilities correspond to a likelihood that the input media is a deepfake.
8. A non-transitory computer readable medium comprising instructions which, when executed, cause at least one processor to at least:
generate a first probability based on a first output of a local binary model manager, a second probability based on a second output of a filter model manager, and a third probability based on a third output of an image quality assessor; and
in response to obtaining a first result and a second result, generate a score indicative of whether input media is authentic based on the first result, the second result, the first probability, the second probability, and the third probability.
9. The non-transitory computer readable medium of claim 8 , wherein the instructions, when executed, cause the at least one processor to determine the input media is authentic when the score is less than a score threshold.
10. The non-transitory computer readable medium of claim 8 , wherein the instructions, when executed, cause the at least one processor to generate the second output by:
applying a Gabor filter to a frame of the input media; and
extracting signals from the frame of the input media responsive to the Gabor filter being applied, the signals corresponding to changes in texture in the input media.
11. The non-transitory computer readable medium of claim 8 , wherein the second result is representative of a degree of blur in frames of the input media.
12. The non-transitory computer readable medium of claim 8 , wherein the first result is a binary value indicative of whether a blink occurred in a human face in the input media.
13. The non-transitory computer readable medium of claim 12 , wherein the instructions, when executed, cause the at least one processor to determine that the blink occurred when a ratio of points in an eye satisfies a threshold for four frames of the input media.
14. The non-transitory computer readable medium of claim 8 , wherein the first, second, and third probabilities correspond to a likelihood that the input media is a deepfake.
15. A method to determine whether input media is authentic, the method comprising:
generating a first probability based on a first output of a local binary model manager, a second probability based on a second output of a filter model manager, and a third probability based on a third output of an image quality assessor; and
in response to obtaining a first result and a second result, generating a score indicative of whether corresponding input media is authentic based on the first result, the second result, the first probability, the second probability, and the third probability.
16. The method of claim 15 , further including determining the input media is authentic when the score is less than a score threshold.
17. The method of claim 15 , further including generating the second output by:
applying a Gabor filter to a frame of the input media; and
extracting signals from the frame of the input media responsive to the Gabor filter being applied, the signals corresponding to changes in texture in the input media.
18. The method of claim 15 wherein the second result is representative of a degree of blur in frames of the input media.
19. The method of claim 15 , wherein the first result is a binary value indicative of whether a blink occurred in a human face in the input media.
20. The method of claim 19 , further including determine that the blink occurred when a ratio of points in an eye satisfies a threshold for four frames of the input media.