Image Annotation
Faceswap Capability Here, we are looking at the following questions while evaluating: Are the identities of the subject in the Face Image preserved in the Target Image? Are there artifacts, limb disfiguration, and/or other logical issues in the images (either Source or Output)? Logical issues: If the source and image are photorealistic, are they looking realistic and having no factual errors? e.g., having a woman's face on a man’s body, skin tone does not match between face and body, the hair does not match Are the expressions in the Source Image and Target Image the same? Expression also includes: smile (showing teeth vs not), direction of eye gaze, eyebrows, and head tilt. Evaluation Instructions: If any of the images contain any of the errors above, select “fail” and choose the failure reason. Otherwise, select “pass”. IMPORTANT: The faceswap should affect exactly one person - the person within the red bounding box in Source Image A If 0 people or more than 1 person are changed