• 

  Recognition of door

• 

  Recognition of kitchen area

• 

  Lama figure

Inspiration

Our primary motivation was exploring how modern AI can be practically applied to enhance accessibility. We wanted to build a tool that doesn't just demonstrate technology, but genuinely helps people by providing a clearer understanding of their environment. ​

What it does

SeeForMe serves as an intelligent visual assistant. It captures visual data from a smartphone camera, processes it to understand the scene, and audibly describes the environment to the user. It bridges the gap between visual input and auditory understanding for the visually impaired. ​

How we built it

We started prototyping the frontend using Vite and standard web tools for rapid iteration. We then integrated the Gemini Flash model to handle the heavy lifting of environmental understanding and connected the ElevenLabs API to give those insights a natural, lifelike voice. The 3D printed mount seen in the demo is currently a concept, but it serves as a blueprint for what users could build themselves or what we could properly manufacture in the future. ​

Challenges we ran into

We faced a significant hurdle accessing the Gemini API. As students, navigating the verification requirement (even for the free trial) was difficult due to the need for personal billing information. We eventually managed to solve the authentication constraints to get the pipeline running. ​

Accomplishments that we're proud of

We are proud to have deployed a functional online application that is ready for use. It’s rewarding to build something that can immediately improve a user's spatial navigation and increase their awareness during daily tasks. ​

What we learned

We learned that AI is incredibly powerful when applied to real-world constraints. Seeing the models work in practice confirmed the potential of multimodal tools in assistive technology. ​

What's next for See For Me

Our roadmap focuses on fabricating the physical concept of the phone mount and further polishing the real-time object detection algorithms to ensure they are robust enough for dynamic environments.

Built With

• elevenlabs
• gemini
• google-cloud
• vite