WeParkit

Inspiration

Every day, UB students, faculty, and staff lose valuable time driving from lot to lot searching for parking. Lots fill quickly, leaving people frustrated and late for classes and work.

The problem isn't just parking availability; it's the lack of real-time information. This inspired WeParkit, a community-driven app where drivers update availability with one tap when they park or leave, helping everyone find a spot faster.

What It Does

WeParkit makes parking simple with just two taps:

  • I'm Parked: GPS automatically identifies which of North Campus's 34 parking lots you're in.
  • I Left: Instantly updates the live map across all users' phones.
  • Space Opening Alerts: Notifies nearby drivers when a parking spot becomes available.

Additional features include:

  • Smart Parking Recommendations: Suggests the best lot, a backup option, and when to leave based on your destination and arrival time.
  • AI Photo Scanning: Uses computer vision to identify open parking spaces.
  • Ice Season Intelligence: Tracks NOAA GOES-19 satellite imagery and NWS weather alerts to anticipate lake-effect snow, adjust parking estimates, and identify icy or inaccessible walkways.
  • Voice Co-Pilot: Lets users ask parking questions hands-free, with spoken responses powered by ElevenLabs.
  • Drive Mode: Provides a full-screen gesture interface for easier, hands-free interaction.

How We Built It

  • Backend: FastAPI and SQLite, with WebSockets for real-time updates.
  • Mobile & Web: Expo React Native.
  • Parking Data: OpenStreetMap lot boundaries, cross-checked against UB campus maps.
  • Computer Vision: OpenCV for parking-space detection, with Gemini as a second opinion.
  • Satellite Intelligence: GOES-19 imagery and optical flow for snow-band tracking.
  • Voice AI: ElevenLabs for speech-to-text and text-to-speech, with Gemini for open-ended queries.
  • Authentication: Gmail sign-in.
  • Reliability: Supervisor process for automatic API recovery.
  • Testing: 119 backend tests, all passing.

Challenges We Ran Into

  • Cold Start: Parking availability depends on community participation. We simplified updates to one tap and added photo scanning to fill information gaps.
  • Map Accuracy: Cross-checking UB parking maps revealed and helped us fix two incorrect lot boundaries and locations.

Accomplishments We're Proud Of

  • Built a complete real-time parking update loop, synchronizing across devices in under a second.
  • Built a conversational voice co-pilot supporting follow-up questions.
  • Clearly labelled simulated data, estimated capacities, and demo scenarios to maintain transparency.
  • Passed all 119 backend tests.

What We Learned

  • Community-driven applications succeed when participation is effortless.
  • Vehicle detection alone doesn't determine available parking; snow, aisles, and stall layouts matter.
  • Transparency about estimated and simulated data builds user trust.
  • Real-world campus networks introduce challenges that don't always appear during local development.

Built With

Share this project:

Updates

Submission history