Illustrative Workshop on Image Generative Models with Dall.E Mini

Content repository for Workshop on Assembling Blocks of Dall.E-Mini and Image Generative Models at ICCIDS '23.

A conference workshop on Image and Video Analysis as part of the IFIP-Springer International Conference on Computational Intelligence for Data Science (ICCIDS), 2023 jointly organized by the Machine Learning Research Group and the Department of CSE, SSN College of Engineering, Chennai, India.

• Link to Illustration Notebooks.
  • Vanilla GAN for Image Regeneration: MNIST data.
  • Autoencoders for Image Regeneration: CIFAR10 data.
  • Variational Autoencoders for Image Regeneration: CIFAR10 data.
  • Vector-Quantized Variational Autoencoders for Image Regeneration: CIFAR10 data.
  • BART for Encoding Text Captions.
  • CLIP to Rank Images by Relevance to Caption.
• Link to Workshop Slides.

Session Overview

• Overview of Dall.E Mini and its Building Blocks
• A Brief Antedate of Autoencoders and Generative Adversarial Networks (GANs) applied in Dall.E Mini
• BART Encoder-Decoder for Image-Text Latent Space Tranlation
• CLIP to Rank Generated Images by Relevance to Captions
• Piecing the blocks of Dall.E Mini together

Implemented with: Python, PyTorch, OpenCV.