Segmenter Variational Autoencoder

November 28, 2021 · View on GitHub

This VAE has the ability to generate from within the distribution of satellites images a new image.

Hyperparameters

HyperparametersValues
Epochs2
batch size32
learning rate1e-4
dropout0.5
mse_weight1
KLDiv_weight1e-3
Hidden activationSeLU
Latent activationSigmoid

The dataset

Data represents satellites images of cities with multiples objects. Capture d’écran 2020-12-03 à 2 22 20 AM

Generative representation of the dataset

The resulted image shows interesting result, where objects are segmented with a specific color which make us able to distinguish an item from another. Capture d’écran 2020-12-03 à 2 31 58 AM