Inspiration
The inspiration behind Digital Eyez came from studying Computer Vision and Computational Photography, where we realized how incredibly complex and valuable human vision is. The ability to perceive depth, recognize objects instantly, and process visual cues is something we often take for granted—yet replicating it computationally is one of the hardest challenges in AI.
AI has made remarkable advancements in mimicking human vision, and we saw an opportunity to use these innovations to empower those who lack or have lost this ability. Having personally experienced vision-related challenges in the past, we deeply understand the frustration of navigating the world with limited sight. This drove us to build a real-time AI-powered assistive tool that can bridge the gap between technology and accessibility, ensuring that visual impairments don’t limit independence or access to information.
What it does
Digital Eyez is an AI-powered assistive tool that helps visually impaired users navigate their surroundings. It provides real-time image analysis, obstacle detection, and text recognition, using GPT-4 Vision. The AI generates context-aware descriptions and converts text to speech, enabling safer, independent mobility.
How we built it
We built our prototype in a collaborative, fast-paced environment. We leveraged version control technology, such as GIT, to allow for efficient and quick collaboration and generative AI tools, such as ChatGPT, to aid in the generation of code.
Challenges we ran into
-Capturing the user's audio input in a concise and accurate manner
-Establishing consistent communication between the front-end and backend
-Trying to run a local model, Ilava , to employ in the image analysis --> had to abandon this idea and switch to OpenAI's API
-Creating a prompt for the model that would assist it in providing a beyond-surface-level analysis of the potential hazards present in an image
Accomplishments that we're proud of
-Speech-to-text parsing of the user's audio input
-Responsive text-to-speech narration of the AI model's analysis of imaged and related prompts
-Having created a functional, self-contained prototype of our idea
What we learned
-How to learn and adapt to new frameworks quickly
-Rapid development and deployment of ideas
-Troubleshooting unexpected behaviors in code
-Collaborative troubleshooting
What's next for Digital Eyez
-Refining our prototype to improve reusability
-Collecting data relevant to obstacles visually impaired people face in day-to-day life to improve models
-Research and development into a light-scale model that can improve scalability
-Testing our ideas in wearables like smartwatches and smart glasses
Log in or sign up for Devpost to join the conversation.