conv_lstm_cell

This repository implements a custom built ConvLSTM cell in Tensorflow and Pytorch.

The model was first introduced in Convolutional LSTM.

The code is not meant to be executable. This code is an outline of how to implement these types of models. For an example of a ConvLSTM that runs see my collision anticipation repo.

Relevant Files

PyTorch implementation of ConvLSTM that initializes weights and defines forward method for inference

see pt_conv2dlstm_cell.py

Tensorflow implementation of ConvLSTM static graph

see tf_conv2dlstm_cell.py

PyTorch implementation of deep neural network containing a ConvLSTM Cell

see pytorch_model.py

Tensorflow implementation of deep neural network containing a ConvLSTM Cell

see tensorflow_model.py

Installing

Packages needed to run the code include:

• numpy
• python3
• PyToch
• Tensorflow