This directory contains the resources relevant to the machine learning aspect of the project. Feel free to follow along the notebooks to learn how to save/convert the models and add metadata, or download the TensorFlow Lite models from TensorFlow Hub.
├── CartoonGAN+ESRGAN.ipynb (an experiment combining two different vision models to see if they produce better quality images)
├── CartoonGAN_TFLite.ipynb (exports TFLite models with dynamic shape support)
├── CartoonGAN_TFLite_Fixed_Shaped.ipynb (this is the main TFLite model export notebook and it exports models with fixed shape)
└── metadata
└── Add_Metadata.ipynb (shows how to populate the TFLite models with relevant metadata)
The white-box CartooGAN TensorFlow Lite models (with metatdata) are available on TensorFlow Hub in three different formats:
- Dynamic-range
- Integer
- float16
This issue contains an informative thread on the conversion process of the original white-box CartoonGAN TensorFlow Lite models.