From the paper I inferred that the model can be trained on custom topology based dataset if I could register provided VOCA raw scans on our topology. Am I correct? or the code is strictly dependent on FLAME topology?
If there any other way to train on custom data?
Hi Timo,
From the paper I inferred that the model can be trained on custom topology based dataset if I could register provided VOCA raw scans on our topology. Am I correct? or the code is strictly dependent on FLAME topology? If there any other way to train on custom data?