DALL-E 2 is a new, multi-billion parameter computer vision model for image generation by OpenAI. It is a successor to the original DALL E, but this time uses a diffusion model. Diffusion models have been gaining traction and are now at the top for image generation. It can also perform tasks like inpainting, upscaling, and interpolation between images. In this video we take a look at the research paper and see how DALL-E has progresses from the original paper to this new DALL-E 2.
Outline:
0:00 - Overview
0:43 - DALL-E 1 Primer
2:16 - DALL-E 2 Overview
4:17 - Example Images & Decoder Discussion
5:45 - DALL-E 2 Training
8:53 - Diffusion Models
9:48 - Experiments
13:27 - My Thoughts
15:01 - Looking At Generated Images
DALL-E 2 blog post: https://openai.com/dall-e-2/
DALL-E 2 paper: https://arxiv.org/abs/2204.06125
Great video on diffusion models: https://www.youtube.com/watch?v=fbLgFrlTnGU