This is inspired by this video by @carykh
In the video, he explains the concept & methods. But lot of things were need to be assume, like the encoders & the LSTM network. It was also worth mentioning it works using MDN. Also the issues like the network repeating the pattern after few seconds(mentioned in carykh's video also), can be seen in video (result.mp4) in this repo.
I used colab for the computation purpose. To work this notebook, open colab & place the weights in correct path in google drive. Adjust the path if needed.
For colab, checkout https://medium.com/deep-learning-turkey/google-colab-free-gpu-tutorial-e113627b9f5d
encoded data can be found here
- carykh youtube video by Cary Huang
- [Time Series Prediction with LSTM Recurrent Neural Networks in Python with Keras](https://machinelearningmastery.com/time-series- prediction-lstm-recurrent-neural-networks-python-keras/)
- Mixture Density Networks by David Ha