Project information

Conditioned diffusion models for image generation

  • Experimented with the hugging face — Diffusers library for conditioned generation of images
  • Utilized Stable Diffusion U-Net model for prediction and the CLIP model to generate textual embedding
  • Decoded the resulting latent image representation using the VAE decoder
  • After implementing Stable diffusion, experimented around and implemented 2 variations - Negative Prompt and Image-to-Image Pipeline