Result: EMFLM2LrP

The ADQ1 method correctly highlights the face region as manipulated.

The BLK method slightly highlights parts in the entire image, and is hence unsuccessful in the manipulation detection.

The CAGI method correctly highlights the face region as manipulated.

The DCT method slightly highlights parts in the entire image, and is hence unsuccessful in the manipulation detection.

In the Comprint fingerprint, we can clearly see a separation of the face and non-face region. In the heatmap, the face is not highlighted in the heatmap, but instead the non-face region is highlighted. This is because the Comprint method does not know which part is fake - it only knows the face region shows inconsistent traces with the rest of the image.

The Noiseprint fingerprint shows a subtle separation of the face and non-face region. In the heatmap, the face is correctly highlighted - successfully detecting the manipulation.

The CATNet method does not highlight any region and is hence unsuccessful in manipulation detection.

Forgery score: 0.988

The TruFor fingerprint (internally using Noiseprint++) shows a subtle separation of the face and non-face region. In the heatmap, the face is correctly highlighted. Additionally, the confidence map is white in the corresponding region, which represents a confident detection. As such, the forgery score is very high (near 1, or 100%). As such, TruFor successfully detects the manipulation.

The FOCAL method correctly highlights the face region as manipulated.