Open zxk19981227 opened 1 year ago
Hi, PyMAF-X can run in real-time but needs the acceleration tools such as TensorRT. Moreover, since the mesh projection needs much fewer points than the original mesh, a sparse SMPL implementation also helps to speed up the execution.
@HongwenZhang Hello, the smplx model is extremly big up to 1G+, is there any plan to train some light-weighted model while might sacrifice some performance?
Could pymaf x run in the realtime?