Inspiration

With all three founders originating from outdoors-heavy backgrounds and major cities, they were fascinated by nature; not only with the freedom it holds but also the dangers it hides. In fact, the National Parks Service and National Geographic have found that day hikers make up 42% of over 46 thousand annual search-and-rescue cases in the USA. Our goal was to utilize the power of the new tech revolution to make sure hiking becomes safer, smarter, and more connected without taking away from the joy it gave all of us!

What it does

TrailMix is an AI-powered hiking safety and mapping platform that is able to automatically detect live hazards, suggest adaptive routes, and lets users come together to share trails with their friends even if offline. The platform utilizes TensorFlow computer detection and Azure-MongoDB cloud integration to identify hazards, assign risk assessments, and dynamically reroute hikers based on a weighted A* algorithm. To add on top of that, it also includes an interactive chatbot to help with safety tips and offline functionality. All in all, hikers should feel more confident even in the most remote locations / pathless areas.

How we built it

-UX Prototyping using Figma to optimize user experience

-TypeScript based frontend to create a mobile/web friendly application

-TensorFlow to allow for real-time hazard detection

-MongoDB to store hazards locally and eventually on the cloud to share dangers

-ElevenLabs to narrate the AI assistant on the trails so they can enjoy the journey

-Azure + OpenAI GPT-4o-mini to allow for risk assessments, route mapping, and emergency scenario handling

-Weighted A* algorithm to dynamically suggest routes based on hazards in the area

Challenges we ran into

-Simulating a nature-based environment in the urban jungle known as Atlanta

-Getting an accurate hazard detection index with a limited dataset

Accomplishments that we're proud of

-Learning several new softwares including TensorFlow and algorithms such as A*

-Integrating a complex backend software and MongoDB to a pleasant frontend

-Staying together as a team over 48 hours of setbacks

What we learned

We learned quickly that the best way to solve a problem is to dive right in. We all had a similar mindset coming in: we love nature and we want to make it better for people. Over 48 hours, we all put in the work to learn new technologies and we have come out as stronger coders but also leaders and friends.

What's next for TrailMix

-Integration with existing hazard datasets to streamline the collection process

-Increasing training of TensorFlow with more object detection datasets

-Expand social aspect of the app and gamify meeting checkpoints

-Allow for better contact of emergency services in desperate situations

Built With

Share this project:

Updates