To address this gap, Ottobock has introduced the #DearAI Community Library, an open dataset designed to provide AI systems with more realistic visual references of people with upper and lower limb prostheses. Microsoft supported the project by providing the technology behind the platform.
More than 60 prosthesis users from around the world contributed images and videos from their everyday lives. Together, they defined what respectful and authentic representation should look like and annotated the selected content for AI training.
The project resulted in two datasets containing around 800 images in total:
- People with upper limb differences
- People with lower limb differences
The datasets are available free of charge on Hugging Face and are intended for researchers, developers, companies, creative professionals and others working with AI.
According to Ottobock, the project follows a simple principle: involving people with disabilities not only as subjects, but as active contributors to the development process. Paralympic swimmer and TV presenter Andreas Onea was among those who helped shape the Community Library by sharing his own experiences.
"If you aren't visible, you don't exist," says Onea. "We want to show AI models that people with disabilities are part of our world and our society. We are involved, we are contributing – and that is exactly what needs to be visible in AI images."
With #DearAI, Ottobock continues its commitment to improving visibility and inclusion, this time focusing on how people with disabilities are represented in generative AI.