Back to Liveportrait

Speed

assets/docs/speed.md

latest902 B
Original Source

Speed

Below are the results of inferring one frame on an RTX 4090 GPU using the native PyTorch framework with torch.compile:

ModelParameters(M)Model Size(MB)Inference(ms)
Appearance Feature Extractor0.843.30.82
Motion Extractor28.121080.84
Spade Generator55.372127.59
Warping Module45.531745.21
Stitching and Retargeting Modules0.232.30.31

Note: The values for the Stitching and Retargeting Modules represent the combined parameter counts and total inference time of three sequential MLP networks.