Attention-based neural machine translation for Multimodal Translation that learns to translate and to predict the visual representation of an image from the encoder.
- Python 2.7
- Theano
- numpy
- lxml
- pexpect
See an example runscript in runscripts/baseline.sh
for how to train a text-only baseline.
See an example in runscripts/multi30kT_multi30kI-InceptionV3.sh
for how to train a model with image prediction.
You need to replace the hard-coded PYTHONPATH variables to the absolute directory on your machine.