BiConvLSTM_Pytorch

Implementation of Bidirectional ConvLSTM in Pytorch

The code from https://github.com/KimUyen/ConvLSTM-Pytorch has been modified to support multiple recurrent layers.

Basic usage

import torch.nn as nn
import torch
from biconvlstm import ConvLSTM

input = torch.rand(2, 5, 3, 64, 64) # (B, T, C, H, W)
gt = torch.rand(2, 5, 64, 64, 64)

convlstm = ConvLSTM((64,64), return_sequence=False, layer_num=2, bidirectional=True)
output, _, _ = convlstm(input)