No rm & riv in the pretrain model BN layers

facebookresearch / video-nonlocal-net

Non-local Neural Networks for Video Classification

Other

1.98k stars 323 forks source link

Open Allen--Wu opened 5 years ago

Allen--Wu commented 5 years ago

Are we sure that running mean and variance are left out for BN layers in resnet C2D pretrain model?

ecr23xx commented 5 years ago

It's might because the BN layer is freeze, so running mean and running variance is not provided.