You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi! Thanks for the excellent work! I just have a simple question about the training code:
With the param --gen_exp_cond and --gen_pose_cond, I think there should be the data dataset_mead.json and dataset_exp.json to be used, although you have checked if there is no json just set the pose and expression to None. So how to get these two jsons? I think the dataset_mead.json is just the dataset.json, what about the dataset_exp.json? Is it the parameter of BFM or FLAME?
The text was updated successfully, but these errors were encountered:
Hi, I encountered the same problem. When I was training, dataset_mead.json was replaced by dataset.json, that is, pose contains 25 dimensional information, but dataset_exp.json didn't know what it was, so it could only be set to None. With the trained model to reenact avatar, there will be two problems, one is that there will be black edges at the bottom, and the other is that the mouth will collapse when the mouth is opened. I don't know if it's because dataset_exp.json is None.
Hi! Thanks for the excellent work! I just have a simple question about the training code:
With the param --gen_exp_cond and --gen_pose_cond, I think there should be the data dataset_mead.json and dataset_exp.json to be used, although you have checked if there is no json just set the pose and expression to None. So how to get these two jsons? I think the dataset_mead.json is just the dataset.json, what about the dataset_exp.json? Is it the parameter of BFM or FLAME?
The text was updated successfully, but these errors were encountered: