Lossy image autoencoders with convolution and deconvolution networks in Tensorflow
This Jupyter notebook refers to: https://www.bonaccorso.eu/2017/07/29/lossy-image-autoencoders-convolution-deconvolution-networks-tensorflow/
Requirements
- Python 2.7-3.5
- Tensorflow
- Keras
- SciPy
- Scikit-Image
- Numba (optional)
Example with CIFAR-10 dataset
(Trained with CIFAR-10 dataset (with 50000 samples) and a code length equal to 128)
|
First row: original images, second row: lossy reconstructions |
Possible improvements
Possible improvements include:
- Adding a flag (using a placeholder) to use the model for both training and prediction. In the former mode, the input is an image batch, while in the latter is a code batch
- Using L1 (and/or L2) code regularization