ConvLSTM

Convolutional LSTM implemented with chainer

python 3.5.2 + chainer 3.0.0

Getting dataset

$ ./get-moving-mnist.sh

$ python3 ./train.py -g 0 --epoch 10 --inf 3 --outf 3 --batch 16

$ mkdir img
$ python3 ./generate.py --model results/model --id 7000 --inf 3 --outf 3

Then, the images are generated in img/.

$ mkdir img
$ wget -O model https://www.dropbox.com/s/pthwoljp8qq3o30/model?dl=0
$ python3 ./generate.py --model model --id 7000 --inf 3 --outf 3

Or, you can try ConvLSTM with the pre-trained model.