ConvRNN in Pytorch

This codes are for Spatial RNN. Both convolutional LSTM and convolutional GRU are implemented in Pytorch. Compuation under GPU/CPU are both supported.

Usage

For LSTM,

import ConvLSTM
layer1 = ConvLSTM.ConvLSTMCell(input_shape=(channel, height, width),
                               hidden_c=hidden_channel_of_LSTM,
                               kernel_shape=(kernel_h, kernel_w)) 

For GRU,

import ConvGRU
layer1 = ConvGRU.ConvGRUCell(input_shape=(channel, height, width),
                             hidden_c=hidden_channel_of_GRU,
                             kernel_shape=(kernel_h, kernel_w)) 

Environment

This codes are tested under Ubuntu 18.04, Python 3.5, Pytorch 0.4.1.