search

MoritzKappel / HF-NHMT

Official PyTorch Implementation of 'High-Fidelity Neural Human Motion Transfer from Monocular Video'
MIT License
85 stars 11 forks source link

Pertained models #3

Closed qo4on closed 3 years ago

qo4on commented 3 years ago

Hi! Did you release pertained models? If not could you tell how much time did it take to train and what hardware?

MoritzKappel commented 3 years ago

Hi! So far, we did not release any pertained models. Depending on the sequence, training took about 3-4 days on a single Titan RTX, and ~20% less on a Quadro RTX 8000 GPU. Also, I recommend loading data from a fast m.2 SSD (we used a Samsung 970 EVO+).