|
发表于 2023-9-18 00:21:47
|
显示全部楼层
DF之外的另一种架构。
DF:Strict face interpretation without morphing. Most accurate and true to source data. Works best when the source and destination have similarly shaped faces and color. Placement of facial features (eyes, mouth, nose, etc.) can differ between source and destination. Works better on frontal shots; possible worse result on profiles. Source faceset must have ample coverage of destination pitch, yaw, and color ranges.
极简描述:更像源脸。
LIAE:Lenient face interpretation with some morphing. Adapts more to destination face. Will adapt to faces with dissimilar shape and color. Placement of facial features may be slightly morphed to fit destination face.
Possible better result on extreme angles and missing data. More replication of destination color and lighting conditions.
极简描述:光影更自然。
|
|