We are playing around with creating a swap using full head training using the BiSeNet FP + hair/ears etc.
Thinking about it - it seems this would only work and look reasonable with subjects with short hair.
If the hair of the subject was to exceed the extracted image - it would look awful - correct?
If anyone on here has experimented with doing full head swaps I'd be interested to hear feedback. Also, how does hair motion appear after converting a video?