Training Motion Net about hidden-two-stream HOT 2 CLOSED

Hi, thank you for your interest.

1. generate_flow_for_ucf101.py is a file for you to generate optical flow offline using MotionNet. It is not for training, but deploying.

2. To train MotionNet, you just need to run something like this:

../../build/tools/caffe train -solver=solver.prototxt

3. For combining two or more caffe models, you can refer to here. Basically, you need three prototxt file, one for model1, one for model2, and one for the new model. Then you load model1, then load model2, and copy the weights to model3, and the save model3. You can search online how to do it, like how to combine two caffe models.

from hidden-two-stream.

Hi Sir,
Thank you so much for your help. That’s really kind of you
I can train the MotionNet now. Im working on merging the two models.

Best,

from hidden-two-stream.