6. Normal AutoEncoder - SummerBigData/Iceberg GitHub Wiki

As we have before, the plan was to train an autoencoder on the entire (labeled and unlabeled) dataset and use the encoded data as the input of my actual classifying neural network.

Here is the model I used:

After the third dense layer is where the encoded image was to be tapped from and fed into a neural network for classification. I attempted to get this working many times, but the output images never did quite look like the input images, just a very generalized depiction of it. I think this is due to how random and noisy the input data is.