Jan. 2026 - Feb. 2026
Synopsis¶
A deep-learning coursework project exploring generative modelling for hot dog image generation. The project involved designing and training a custom Variational Autoencoder, integrating the pretrained Stable Diffusion VAE, and implementing a U-Net-based Denoising Diffusion Probabilistic Model that generates images within a compressed latent space.
Skills¶
Below are the skills and experiences gained from the project:
Knowledge of generative modelling principles and latent representations
Practical experience designing and training Variational Autoencoders
Practical experience implementing and training diffusion models
Competency with PyTorch and deep-learning development workflows
Experience evaluating and visualising generated images and learned latent spaces
