2D Convolutional Long-Short Term Memory (LSTM) layer. Similar to a normal LSTM, but the input and recurrent transformations are both convolutional. Corresponds to the ConvLSTM2D Keras layer .
- Type: Keras Deep Learning NetworkKeras NetworkThe Keras deep learning network to which to add an ConvLSTM2D layer. The shape of the tensor must be [time, height, width, channel] or [time, channel, height, width] for data format channels_last and channels_first respectively.
- Type: PortObjectKeras NetworkAn optional Keras deep learning network providing the first initial state for this ConvLSTM2D layer. Note that if this port is connected, you also have to connect the second hidden state port. The shape must be [height, width, channel] or [channel, height, width] depending on data format and the dimensionality of the channel dimension must match the number of filters of this layer.
- Type: PortObjectKeras Network> An optional Keras deep learning network providing the second initial state for this ConvLSTM2D layer. Note that if this port is connected, you also have to connect the first hidden state port. The shape must be [height, width, channel] or [channel, height, width] depending on data format and the dimensionality of the channel dimension must match the number of filters of this layer.