Experimenting with Autoencoding Architectures to Encode World State Representations
This is a small framework for training and testing auto-encoding architectures for the purpose of encoding world state representations into latent space and thus efficiently compressing the essential world properties depicted in the training image dataset.
Clone repository to a desired directory:
cd C:\dev\
git clone https://github.com/blazm/baxter.git
Install Anaconda with Python >= 3.5.
Create virtual environment with the following modules (the code might work with > versions):
numpy == 1.16.4 scipy == 1.1.0 cv2 == 4.1.0 tensorflow == 1.2.0 Keras == 2.0.7 Pillow
File config.ini includes all parameters to train the models.
Setup the parameters, then run training:
activate conda_env
(conda_env) python train_ae.py
Results will be saved in the snapshots folder (.pdf, .csv, .ini) and in tf-log folder (TensorBoard training log).
Blaz Meden – @blazm – [email protected]
Distributed under the XYZ license. See LICENSE for more information.