Documentation Summary¶
This documentation gives a overview of the code used to implement and train the architectures presented in this master’s thesis. The code has several dependencies as listed below:
Dependencies¶
- TensorFlow
- TfLearn
- Numpy
- scikit-learn
- pickle
- glob
- scipy
In the current state, most of the code is not runnable for verification. This is because the code also depends on pretrained models to be able to train the Spatial Stream and the Recurrent Networks. It is however possible to run the Temporal Stream, but since our dataset is not yet made openly available, the results from this run will only be to verify that the code can run.