search

loliverhennigh / Convolutional-LSTM-in-Tensorflow

An implementation of convolutional lstms in tensorflow. The code is written in the same style as the basiclstmcell function in tensorflow
Apache License 2.0
420 stars 163 forks source link

Structure of network #8

Open deckyal opened 6 years ago

deckyal commented 6 years ago

Hi, just a quick question. How is the structure of the network in the main_conv_lstm.py??

If I understand correctly is that the network passed the image to the 4 layer of convolutional. Then to one layer of conv lstm. Then to 4 layer deconvolutional. Isn't it?

If yes then, how to use multi layer of this conv lstm?.

Thanks

loliverhennigh commented 6 years ago

That is what is happening with the network architecture. If you want a better architecture consider checking out the branch here https://github.com/loliverhennigh/Convolutional-LSTM-in-Tensorflow/tree/residual_model . It uses residual connections and stuff. I wrote it as an example for someone a while back. It should kinda give you an example of how to do multi layer conv lstm. Otherwise ConvLSTMs are now in TensorFlow master and can be used as normal recurrent networks. Here is an example of using them with dynamic_rnn method https://github.com/loliverhennigh/dynamic_rnn_conv_lstm/blob/master/mnist_deep.py#L65. It should be pretty easy to get them to work in a multi layer way with that.