search

yjxiong / caffe

A fork of Caffe with OpenMPI-based Multi-GPU (mainly data parallel) support for action recognition and more. More documentation please see the original readme.
http://caffe.berkeleyvision.org/
Other
551 stars 154 forks source link

Single-frame FlowNet #43

Closed mathrho closed 8 years ago

mathrho commented 8 years ago

I'm wondering have you ever tried to train the VGG16 temporal model with single-frame, i.e., length=1 instead of length=10? It seems that initialize the model by pre-trained RGB model from ImageNet does not help if I set length=1 when train the temporal flow model.

wanglimin commented 8 years ago

I am sorry that we did not try single frame ConvNet for optical flow.