Henna, or mehndi, is an ancient art of temporary body adornment, deeply rooted in cultural celebrations, rituals, and personal expression. This project seeks to honor and extend this tradition by leveraging machine learning to create new and unique henna designs and introduce the artform to a wider audience in an accessible and engaging manner. We were inspired to explore the intersection of generative AI, diffusion models, traditional art forms, and the rich tapestry of South Asian culture.

By fine-tuning a model on a vast dataset of historical and contemporary henna patterns, we aim to not only generate novel designs but also to provide a platform for cultural exploration. The aim is to democratize access to this beautiful art form, allowing users to engage with and learn about its significance, while simultaneously pushing the boundaries of what's possible with diffiusion models. We find inspration in the beautiful ad-hoc designs done at pre-wedding parties -- and just as much in the ever-improving tooling at Huggingface, Replicate, and DeepMind.

This project is further motivated by the exciting possibilities within the field of machine learning model operations (MLOps). We use this project to explore the ever-improving tools and platforms available for fine-tuning and deploying generative AI models in cost-effective ways. Think of it as a bridge between the past and the future, between tradition and innovation, offering a unique blend of cultural appreciation and technological advancement.