A convolutional approach to estimate surface geometry for photo-realistic rendering
Motion transfer booth for a 1 hour everybody dance now video generation using EdgeTPU and Tensorflow 2.0
