SafeVis

Inspiration

Leveraging AI models to enhance the sensory perception of individuals with vision impairment and hearing impairment enables them to better understand their environment and enhances their

What it does

Employing a custom object detection model, the system identifies and alerts individuals about potential hazards, such as approaching vehicles or potential threats like stalkers.

How we built it

Download a pre-trained YOLOv8 model and further train it to improve the detection of moving vehicles. Develop an algorithm to estimate the level of danger posed by these vehicles, and deploy the system for practical use.

Challenges we ran into

Adapting the most recent version of PyTorch to CoreML presents a significant challenge, requiring exploration of various libraries to ensure the smooth execution of both the model and associated scripts locally and on-device.

Accomplishments that we're proud of

We take pride in our achievement of successfully deploying the solution to run directly on the device, eliminating the need for reliance on external APIs. This enhances efficiency, reduces latency, and provides users with a seamless experience.

What we learned

During this project, we gained valuable insights into the challenges of training custom models and deploying functional applications across different platforms in a short timeframe. Despite the difficulties, we deepened our understanding of machine learning and developed numerous custom algorithms to interpret and utilize the AI outputs effectively for various position computations.

What's next for SafeVis

Moving forward, our next steps for SafeVis involve advancing this proof of concept to further assist individuals with vision and hearing impairments in increasing their awareness of their surroundings. We aim to refine our technology, enhance its capabilities, and explore additional features to provide even greater support and empowerment to those who rely on it. This includes conducting further research, gathering user feedback, and iterating on the design to ensure it meets the diverse needs of our target users. Additionally, we plan to explore partnerships and collaborations to broaden the impact of SafeVis and make it more widely accessible to those who can benefit from it.