ParkingPercent – AI-Powered Parking Lot Analytics Platform

Multi-platform SaaS application that automates parking lot utilization tracking using computer vision to provide real-time occupancy analytics for property managers and facility operators.

The Problem

Property managers and facility operators need to understand parking lot utilization patterns to optimize space allocation, justify expansions, and improve tenant satisfaction. Manual counting is impractical, and traditional sensors are expensive to install and maintain.

The Solution

ParkingPercent enables automated parking analytics by integrating directly with existing security camera infrastructure. Cameras send images via REST API at scheduled intervals (hourly, daily, etc.), where our Detectron2 computer vision models automatically detect and count vehicles, calculates occupancy rates, and tracks usage patterns over time—leveraging cameras already in place without additional hardware.

Key Technical Achievements

API-First Architecture

- Security Camera Integration: REST API designed for automated image uploads from existing CCTV/IP camera systems

- Scheduled Snapshots: Cameras configured to POST images at intervals (e.g., every hour) for continuous monitoring

- Webhook Callbacks: Real-time processing status updates sent back to client systems with detection results

- Unkey Authentication: Secure API key management for camera systems and third-party integrations

- Web Dashboard Alternative: Browser-based upload interface for manual snapshots, testing, or systems without API access

Photo-to-Insight Pipeline

- Direct Upload Flow: S3 presigned URLs for client-side uploads → GPU-accelerated processing → real-time webhook callbacks

- EXIF Metadata Extraction: Automatically extracts timestamps and camera metadata for accurate time-series tracking

- Timezone Intelligence: Uses geo-tz library with GPS coordinates (or manual configuration) for accurate time-series analytics

- Thumbnail Generation: Automatic thumbnail creation with Sharp for fast dashboard loading

Computer Vision Pipeline

- YOLOv8 Object Detection: Custom-trained model for vehicle detection in parking lots with 95%+ accuracy

- GPU-Accelerated Processing: Modal.com serverless infrastructure with NVIDIA T4 GPUs for sub-30-second processing

- Bounding Box Storage: PostgreSQL storage of detection coordinates for visual overlay and verification

- Scalable Inference: Handles multiple simultaneous uploads from different camera feeds

Analytics & Insights

- Time-Series Tracking: Historical occupancy trends with date range filtering and timezone-aware display

- Occupancy Rate Calculation: Automatic calculation of utilization percentage based on total spots vs. occupied spaces

- Interactive Visualization: Recharts-powered graphs showing peak usage times, daily patterns, and trends

- Multi-Lot Management: Support for multiple parking lots per account with individual analytics

Full-Stack Implementation

- Backend: PostgreSQL (Prisma ORM), Redis rate limiting, AWS S3 for image storage with automatic thumbnail generation

- API Security: Unkey authentication for API keys, per-endpoint rate limiting (20 uploads/min), storage quota enforcement (10GB per user)

- Frontend: Next.js 15 with React 19, responsive dashboard, Konva canvas for detection overlays

- Auth & Payments: Clerk authentication for web users, Lemon Squeezy subscription billing (Merchant of Record)

- Webhook System: Real-time processing callbacks to client systems with detection data

DevOps & Testing

- Comprehensive Test Suite: Unit, integration, and security tests with parallel CI/CD execution

- Local Webhook Development: Cloudflare Tunnel setup for testing processing callbacks locally

- Strict Type Safety: TypeScript with zero implicit any, enforced via ESLint and CI

Tech Stack

Next.js 15, TypeScript, PostgreSQL, Redis, AWS S3, Modal.com, YOLOv8, Clerk, Lemon Squeezy, Unkey, Prisma, Tailwind CSS v4, Bun, Sharp (image processing), EXIFR (metadata extraction)

Outcomes

- Production-Ready: Deployed with full API, webhook callbacks, payment processing, and analytics dashboard

- Leverages Existing Infrastructure: Uses security cameras already installed—no new hardware required

- Scalable Architecture: Serverless GPU processing handles multiple camera feeds without infrastructure management

- Developer-Friendly: RESTful API enables seamless integration with camera systems and property management software

- Cost-Effective Solution: Eliminates need for dedicated parking sensors while providing automated, continuous monitoring