Peer Review Presentation

This week, our team got to work on our presentation for the preliminary design report. Our slides came out fantastic and we had the opportunity to practice presenting in front of our peers and receive feedback. The team did great during the presentation, but some of the feedback we received was mostly to practice more to memorize the information we need to present. For the upcoming week, we are excited to present our PDR to Fanatics on Wednesday and show off all our hard work and research!

Stable Diffusion XL

Upon further research, our team discovered this version of stable diffusion that has more features. It improves on the original in three different ways:

• The UNet is 3x larger and SDXL combines a second text encoder (OpenCLIP ViT-bigG/14) with the original text encoder to significantly increase the number of parameters

• Introduces size and crop-conditioning to preserve training data from being discarded and gain more control over how a generated image should be cropped

• Introduces a two-stage model process; the base model (can also be run as a standalone model) generates an image as an input to the refiner model which adds additional high-quality detail

Plans For Next Week

• Update our software architecture to embed into Fanatics architecture
• Improve on our presentation based on feedback we received and prepare
• Continue learning about Stable Diffusion XL