Around Me

• 

  Sleeve display

Inspiration

Every year our devices become a more natural extension of ourselves, and our team was inspired to look 5 years into the future, engineering a device that takes advantage of the latest DL CV models, Mobile Computing, and Distributed Cloud Databases. Real-time 360° person and object detection provides services that were previously unavailable (such as automated inventory management, consumer traffic analysis, traffic route planning, etc.). Essentially, looking forward to futuristic, smart, advanced cities with wearable/mobile AI combined with powerful edge computing motivated our idea.

What it does

_ Around Me _ is a wearable device and object/person detection platform. On the user's head is a rotating camera that constantly records their surroundings and performs object detection. Each frame, their surrounding bounding boxes are saved, and the positions of surrounding people and various objects are displayed on the user's sleeve which has a screen and a built-in mobile computer. Additionally, all the detected bounding boxes are stored in a cloud-hosted Redis database and visualized for quick access to real-time accumulating perception data of the user's nearby surroundings.

How we built it

The project is divided into 2 parts: cloud database & edge device. The edge device was an RPi 4B with an LCD screen, webcam, and motor. This is where ML inference was run, and the cloud portion was where the perceived objects were uploaded (to the Redis DB) for data visualization and analysis.

Challenges we ran into

Some of the challenges we ran into were: Power delivery to all components used in the system Calibrating the GUI to properly display the information given from the ML model ML inference latency ono RPi 4B

Accomplishments that we're proud of

Integrating the entire hardware system with cloud integration.

What we learned

Redis, using Object Detection Models, Pytorch, Raspberry Pi, PyGame, MatPlotLib

What's Next for Around Me

Expanding the cloud platform to synthesize multiple user's data to construct useful insights such as forensic analysis, and consumer traffic.

Built With

• ai
• computer
• detection
• machine-learning
• matplotlib
• object
• python
• pytorch
• raspberry-pi
• redis
• smartcity
• vision
• yolo