SafeSteps

Inspiration

We were inspired to create Safe Steps after witnessing the challenges faced by elderly individuals and their caregivers in ensuring safety and well-being in care homes. We recognized the need for a reliable and efficient solution to detect falls and provide prompt assistance.

What it does

Safe Steps is an AI-powered software that detects falls in elderly care homes. It allows caregivers to broadcast their cameras, which a local model then analyzes to identify potential falls. When a fall is detected, a snippet contacting frames with the fall is sent to the Google Gemini API for confirmation, ensuring accurate fall detection and prompt response.

How we built it

We built Safe Steps using a combination of local AI models and the Google Gemini API. The local model was forked from this link. We integrated the local model with a user-friendly interface that allows caregivers to broadcast their cameras easily. When a potential fall is detected, the video is securely transmitted to the Google Gemini API for further verification.

Challenges we ran into

Time. The biggest challenge we faced was time. I'm currently a student and was forced to take a couple of weeks off to lock in and get ahead so I would have time to finish this project. Outside of that, I had no idea how hard it would be to stream a video to and from an endpoint.

Accomplishments that we're proud of

The biggest accomplishment is really a proof of concept. This is a local project we built on our laptop to demonstrate how it could be done. I, unfortunately, don't have the time or resources to host this model locally or quantify it to reduce computational costs, but just putting the idea out there, showing the feasibility of it, and someone possibly using it in the future to save lives is amazing to me.

What we learned

One crucial lesson learned from this project was the significance of consistent development environments. Collaborating across different platforms, such as Mac and Windows, can pose significant challenges and hinder progress. This became evident when an early team member struggled to set up the local model on their machine. This led to their unfortunate departure from the project, emphasizing the need for a unified development process.