Inspiration
We wanted to create a project that would have a sustainable impact and would include all our group members to contribute, so after much deliberation, we decided to create this project - Midnight Walker
What it does
This project is a proof of concept that by integrating object detection to street lights, such that the lights only are bright when a person walks within the field of range, we can radically reduce the amount of energy that is consumed by the street lights.
How we built it
We built this app using TensorFlow.js, COCO-SSD Model, and WebRTC API for the main functionality of the object detection. Then using the HTML, CSS, and JavaScript, we built the rest of the website.
Challenges we ran into
The object detection model provided a wealth of data including position coordinates, bounding box dimensions, and confidence scores. We had to determine which data points would be most relevant for controlling lighting effectively. Initially we struggled with how to translate detection coordinates into meaningful lighting controls.
Accomplishments that we're proud of
That we were able to finally have an MVP.
What we learned
Technical Skills:
- Hands-on experience with TensorFlow.js and machine learning implementation
- Real-time object detection integration in web applications
- Camera API integration and handling browser permissions ## What's next for MidNight Walker To create a physical prototype.