search

neeek2303 / EMOPortraits

Official implementation of EMOPortraits: Emotion-enhanced Multimodal One-shot Head Avatars
311 stars 17 forks source link

Confusion about ExpressionEmbed #18

Open chenerg opened 1 month ago

chenerg commented 1 month ago

In your paper, there exists an expression model for predicting R, T, and expression vector(Z_s). However, in the expression_embedder.py file, I only find variables such as thetaand eye_vector. Could you please inform me which variable is related to the expression vector as mentioned in the paper?

johndpope commented 2 weeks ago

should be 

     self.net_head = ResNetWrapper - backbone50

   data_dict['theta'], params = point_transforms.estimate_transform_from_keypoints(data_dict['keypoints'], self.aligned_keypoints)
        data_dict['scale'], data_dict['rotation'], data_dict['translation'] = params