PredNet implementation of PyTorch.
Environments are in env.yaml.
The network could be found in prednet.py which importing convlstmcell.py.
Train in train.ipynb, and test in test.ipynb.
In folder named 'KITTI', you can find the samples of datas used to testing this Predet.
Save processed datas in kitti_data folder as
for training : X_train.hkl, sources_train.hkl
for validation : X_val.hkl, sources_val.hkl
for test : X_test.hkl, sources_test.hkl
(you can download them in here)
Datas loaded with kitti_data_load.py when running train.ipynb.
Code and models accompanying Deep Predictive Coding Networks for Video Prediction and Unsupervised Learning by Bill Lotter, Gabriel Kreiman, and David Cox.
The PredNet is a deep recurrent convolutional neural network that is inspired by the neuroscience concept of predictive coding (Rao and Ballard, 1999; Friston, 2005). Check out example prediction videos here.
- https://coxlab.github.io/prednet/
- https://github.com/leido/pytorch-prednet
- https://github.com/jonizhong/afa_prednet
The name of files like jupyter_ or colab_ is for indication.
You have to remove them before use it.
And always be careful DIRECTORIES.
