Huertas97/GenAI-FashionMNIST · Apply for community grant: Academic project (gpu)

Dear HF managers,

This HF space is a Gradient demo of the capabilities of a Class-Conditiones UNet Diffusion model for generation of images from the Fashion-MNIST dataset.

This project aims to develop and deploy class-conditional image generation models using the Fashion MNIST dataset. Unlike typical datasets like MNIST, Fashion MNIST offers more complex and varied clothing images, presenting unique challenges and opportunities for research.

The model (safetensors weight and noise scheduler), code, and documentation are available in HF models and in GitHub.

A GPU grant will enable faster inference throught the denoising process (4000 steps, 20 s in a GPU), allowing users to interactively generate images and experiment with different class conditions on our Hugging Face Space.