Stable Diffusion Hackathon summary: the winner’s story
Voices of to the lablab community!
We, the lablab.ai, want to provide best tools for developing AI based aps. During our AI Hackathons you can get assistance from our mentors, access to the hottest AI based technology from our partners, support from an amazing community of builders, creators and innovators and everything else which is required for our community to build during our generative AI Hackathons. Also, after the event is finished, winners are chosen and dust starts to settle in, we don’t stop working - we continue to work hard by helping participants to present their project during our dedicated twitch streams, individual talks with potential investors and giving place to talk about their experience and road to that very moment. And in this article we give voice to Rahel Gunaratne, winner of Transformers and Stable Diffusion AI Hackathon!
I found lablab.ai through Instagram. I was mindlessly scrolling as one does on those addictive sites. I immediately messaged my brother, Sacha Gunaratne who is already very familiar with neural net architectures, and I asked him if he would join me. He said yes.
He and I work very well in a team. Both tackling tasks in parallel but still solving each other’s problems. We try to play to our strengths. For example, he was familiar with Streamlit so he set that up while I collected the data from arXiv. I enjoyed working with Sacha because we understand each other very well and we are always willing to listen and implement each other’s ideas. But none of it could have happened without LabLab.
LabLab supports us with infrastructure to make these projects possible. For example, during the Stable Diffusion Hackathon they provided an image generator API endpoint for participants to use. In addition, they provide many tutorials in their discord/website. They understand what is required for the hackathon projects and don’t just expect participants to already know everything. I had never used a Transformer model or a Stable Diffusion model before the hackathons and now I can say that I have used both in a practical sense with great success.
I think the topics that LabLab have chosen for their hackathons are extremely important as well. Both Stable Diffusion and Transformers are novel technologies. So, they are very applicable in a business sense because many companies have yet to use them. This means there can be many new products on the market. Not just improved existing ones, but completely new. It means your project, if it uses Stable Diffusion or Transformers, can probably be turned into a product, and dominate the non-existing market for it. Apart from a business sense, I think this technology has the potential to make peoples’ lives much easier. From text summarization to image generation, the possibilities are endless.
I won’t talk too much about the projects I worked on during the Hackathon. You can view the project page for a better understanding of each one.
Transformers AI Online Hackathon winning project
One of the projects that I worked on was a research paper semantic search/clustering. It helps researchers query, find similar papers, or identify similar concepts in completely different fields. Transformers can identify connections on a large scale and to a high degree of accuracy. Concepts that apply to one field can be translated into completely different fields. For example, Factor Analysis (a computer science concept) was used to solve the problem of discovering connections in gene networks.
Stable Diffusion AI Hackathon winning project
The other project I worked on was a Stable Diffusion workflow tool. It is used to interface with Stable Diffusion models to help users create videos from specific images/prompts. It can be hard to get what you want from Stable Diffusion. You must do prompt engineering or even model finetuning. Both of which are currently inaccessible to the average person. With this tool, you can generate the images as key frames. Then you can interpolate them to create a video that will show the images you selected.
All in all, thank you LabLab.ai. You have given us a great opportunity and I am certainly grateful for it.
We would like to thank all the participants of our AI Hackathons, for their dedication in delivering the highest quality models, hard work to create new solutions to world's problems and passion to work with AI!
You, as a community, show that with an AI there are no limits and everyone can start their journey with generative AI technology, no matter their previous field of expertise!
And we would like to invite all of you to join our upcoming AI Hackathons and build an amazing solution to world's problems in just 7 days!