devHacks2026 — Crowd Counter

Real-time people counting system built for bars and venues to track occupancy and stay within fire safety limits. Uses a camera pointed at a doorway with a virtual tripwire line, when someone crosses one direction the count goes up, the other way it goes down.

Built by Ilia Mazor and William Gore.

Why

We're both bartenders and recently the LGCA came by our workplace. They mentioned they still manually count people to check fire code compliance — that's slow, error-prone, and easy to lose track of during a busy night. We wanted to automate it with a camera and some computer vision.

How it works

• A camera watches a doorway
• YOLOv8 detects people in each frame
• ByteTrack follows each person across frames
• A tripwire line is drawn across the entrance — when someone's bounding box center crosses it, we count them in or out
• The count is sent to a central server and displayed on a web dashboard

Setup

pip install -r requirements.txt

Requires Python 3.11+. First run will download the YOLOv8n model (~6MB).

Quick start

The easiest way to get everything running:

Windows:

start.bat

Mac/Linux:

chmod +x start.sh
./start.sh

This starts the coordinator server and the web dashboard. Then open http://127.0.0.1:8001 in your browser to access the dashboard.

You can also start things manually:

1. Start the coordinator server

python coordinator/cameraServer.py

Optionally pass a capacity limit:

python coordinator/cameraServer.py 50

2. Start the web dashboard

python dashboard/webServer.py

3. Add a camera

You can add cameras through the web dashboard — click two points on the canvas to place the tripwire line, set your source and camera ID, then hit "Add Camera".

Or run the detector directly:

python node/detector.py

CLI options

Argument	Default	Description
--source	0	Camera index or path to a video file
--camera-id	entrance_cam	Unique name for this camera
--line-start	320 50	Tripwire line start point (x y)
--line-end	320 430	Tripwire line end point (x y)
--calibrate	off	Click to place the tripwire line visually
--flip	off	Swap in/out direction for exit cameras
--brightness	0	Brightness boost (0=off, 1=light, 2=moderate, 3=max)
--stream-port	0	Port for MJPEG live stream (0=disabled)

Calibration

Use --calibrate to place the tripwire line on a live feed:

python node/detector.py --calibrate

Click two points, press r to reset, press q when done. The terminal prints the exact args to reuse next time.

You can also set the line through the web dashboard canvas — no CLI needed.

Direction flipping

By default crossing left-to-right counts as "in". For exit cameras use --flip:

python node/detector.py --flip --camera-id "exit_cam"

Dashboard

The web dashboard at http://127.0.0.1:8001 lets you:

• See the current count in real time
• Set the venue capacity
• Add and remove cameras
• Draw the tripwire line on a canvas (no need to guess coordinates)
• View live camera feeds via MJPEG streams

Architecture

Camera Node (detector.py)          Coordinator (cameraServer.py)
+---------------------+            +----------------------+
| Webcam/Video        |            | Port 8069 (cameras)  |
|   -> YOLOv8n        |  crossing  |   -> updates COUNT   |
|   -> ByteTrack      | ---------> |                      |
|   -> Tripwire line  |            | Port 8067 (web)      |
|   -> MJPEG stream   |            |   -> serves data     |
+---------------------+            +----------+-----------+
                                               |
                                   +-----------+-----------+
                                   | Dashboard (webServer) |
                                   | Port 8001             |
                                   |   -> serves UI        |
                                   |   -> proxies to coord |
                                   +-----------------------+

Project structure

devHacks2026/
├── node/
│   ├── detector.py      # Camera node — detection, tracking, tripwire
│   ├── client.py        # Sends crossing events to coordinator
│   └── config.py        # Default configuration
├── coordinator/
│   └── cameraServer.py  # Central server — tracks count
├── dashboard/
│   ├── webServer.py     # Serves the web UI
│   ├── index.html       # Dashboard page
│   ├── script.js        # Dashboard logic + calibration canvas
│   └── style.css        # Styling
├── start.bat            # Start all servers (Windows)
├── start.sh             # Start all servers (Mac/Linux)
├── requirements.txt
└── README.md

Future improvements

• Support for multiple cameras covering different entrances/exits
• Person re-identification to avoid double counting across cameras
• Persistent storage so counts survive server restarts
• Push notifications or alerts when approaching capacity
• Historical data and analytics for tracking peak hours
• Better handling of crowds and groups crossing at the same time
• Cloud-hosted dashboard for remote monitoring
• Authentication on the dashboard