facebookresearch / InterHand2.6M

Official PyTorch implementation of "InterHand2.6M: A Dataset and Baseline for 3D Interacting Hand Pose Estimation from a Single RGB Image", ECCV 2020
Other
676 stars 92 forks source link

theTraining speed is slower than yours paper #86

Open kidSun opened 2 years ago

kidSun commented 2 years ago

Hello,I use four nvidia 2080ti GPUs to train the 5fps datasest, but need 6.47h/epoch, I see you train your model for 20 epochs with four NVIDIA TitanV GPUs, Is this normal? image

mks0601 commented 2 years ago

Yes due to the huge size of the dataset :( As suggested in the dataset homepage, I recommend training and testing on human annotation subset.

kidSun commented 2 years ago

ok,thanks you