ParkingLotRevenuePredictor

🅿️ Parking Lot Revenue Predictor

Parking Lot Occupancy → Store Revenue Prediction
Using Computer Vision + Time Series Forecasting to estimate store earnings from parking lot images.
Built for a hackathon, powered by Python, OpenCV, Streamlit, and Prophet.

---

📌 Project Overview

Retail chains like Walmart or Costco often see a direct correlation between parking lot traffic and store sales.
This project uses parking lot images (from the PKLot dataset) to:

1. Detect cars using computer vision.
2. Estimate customer traffic from parking lot occupancy.
3. Forecast store revenue using time series models.

💡 Hackathon Edge:

• Unique bridge between computer vision and economic forecasting.
• Uses a real-world dataset (PKLot) for realistic results.
• Deployable via Docker for instant demonstration.

---

🛠 Tech Stack

Core

• Python 3.12
• OpenCV → Car detection from parking lot images
• Prophet → Revenue forecasting
• Pandas / NumPy → Data manipulation

Visualization & UI

• Streamlit → Interactive dashboard

Deployment

• Docker → Containerized app
• docker-compose → Multi-service orchestration

---

📂 Project Structure

ParkingLotRevenuePredictor/
│
├── traffic_data_builder.py # Builds dataset from images
├── car_counter.py # Counts cars in images using OpenCV
├── revenue_forecast.py # Forecasts revenue from traffic data
├── dashboard.py # Streamlit app for visualization
│
├── requirements.txt # Python dependencies
├── Dockerfile # Docker build config
├── docker-compose.yml # Compose configuration
└── README.md # Project documentation

---

1. Add Your Dataset Place your PKLot images in:

data/

1. 🐳 Run with Docker Build and Run:

docker build -t parking-lot-revenue .
docker run -p 8501:8501 parking-lot-revenue

1. 🐳 Run with docker-compose

docker-compose up --build

Then open the app at: ➡ http://localhost:8501

💡 Example Prediction Flow Upload PKLot dataset images into the data folder.

The app counts cars in each image.

Maps the number of cars to estimated sales figures.

Uses historical data to predict next week/month's revenue.

📊 Workflow

1. Data Input → PKLot dataset images (JPG/PNG).
2. Car Detection → OpenCV detects and counts cars.
3. Traffic Dataset Creation → Image timestamps + car counts → traffic CSV.
4. Revenue Mapping → Car counts mapped to estimated revenue.
5. Forecasting → Prophet predicts future revenue trends.
6. Visualization → Streamlit dashboard displays occupancy, traffic, and revenue forecast.

---

🚀 Live Demo

🎯 Try it instantly on Hugging Face Spaces — no install needed:
➡ Run the App

---

⚠️ Limitations

• PKLot dataset is from Brazil, so direct sales mapping may need calibration for other regions.
• Lighting/visibility in images impacts detection accuracy.

---

Note:
This repository was originally created on August 15, 2025, prior to the release of the Bay2Bay Hacks guidelines.
While the initial dataset setup and some foundational scripts were prepared earlier,
all hackathon-related development, improvements, and final deployment have been done during the event period.