Currently, we provide the following models:

• rnnl1: 1-hidden-layer RNN
• rnnl2: 2-hidden-layer RNN
• rnnl1gc: 1-hidden-layer RNN with grammar-cell mapping units as input
• rnnl2gc: 2-hidden-layer RNN with grammar-cell mapping units as input

change the path_to_project_folder to the folder you store optical-flow-pred in

• scripts/*.sh
• all the hyperParam.py in the examples you will run

1. Bouncing Balls

cd data/bouncing_balls_generator
bash run.sh

cd examples/name_of_model
bash run.sh

After the training, you could visualize the prediction results by:

bash visualize.sh