ParkALot

• 

  Algorithm identifies an open space as a car leaves

• 

  A text message alerting the user of an open space is promptly sent with Twilio API

Inspiration

I commute to class every day by car and often struggle to find parking. This problem prompted me to think of ways to find parking more easily.

What it does

The program uses computer vision on videos from security cameras, for instance, to identify open spots to park. It then notifies the user via SMS.

How I built it

I used a R-CNN trained on the COCO dataset to recognize parked cars, determining empty spots when parked cars would leave.

Challenges I ran into

The speed of the Network is very slow, in part due to my computer's lack of processing power. I struggled to set up my model on a Google Cloud Virtual Machine.

Accomplishments that I'm proud of

The model correctly identifies when a spot vacates and alerts the user promptly.

What I learned

I learned how to use OpenCV in conjunction with predictive models to process videos.

What's next for ParkALot

Speed improvements are definitely needed, including optimizations for different camera angles. Map integration and UI is also needed.

Built With

• keras
• numpy
• opencv
• python
• tensorflow