virlus / vae

AI3611 VAE project

Reproducing the best model

The best model can be trained and evaluated through python main.py

Inference results

latent_1_eval and latent_2_eval store the inference figures by latent dimensions of 1 and 2 respectively, where the names of figures stand for the value of latent vectors.

Training results

results_* directories store the best models for a certain number of latent dimensions and the corresponding evaluation results throughout the whole training process.

Inference script

Running evaluation.py gives you inference figures.