Name		Name	Last commit message	Last commit date
parent directory ..
kitti		kitti
.DS_Store		.DS_Store
README.md		README.md
evaluate.py		evaluate.py
frame_prediction.ipynb		frame_prediction.ipynb
frame_prediction_audio.ipynb		frame_prediction_audio.ipynb
frame_prediction_ucf.ipynb		frame_prediction_ucf.ipynb
keras_utils.py		keras_utils.py
prednet.py		prednet.py
prednet_model.py		prednet_model.py
scratch.ipynb		scratch.ipynb
settings.py		settings.py
train.py		train.py
utils.py		utils.py
visualization.ipynb		visualization.ipynb

README.md

Predictive Coding model

This contains scripts to train and extract from Lotter et al. predictive coding model.

evaluate.py: script to extract predictive coding features and generate frame predictions.

Example: extract features using a PredNet with random weights for data from 3 classes of the Moments in Time dataset (defined in settings.py)

> python evaluate.py prednet_random_finetuned_moments__representation --task 3c

train.py: script to train the predictive coding neural network. Example: train PredNet from scratch using data from 10 classes of the Moments in Time dataset (defined in settings.py)

> python train.py prednet_random_finetuned_moments__representation --task 10c