I've always been fascinated by computer vision, but the existing tools always felt like a compromise. You either pay exorbitant monthly fees for enterprise software, or you hand over your video feeds to a cloud server you don't control.
That didn't sit right with me. I wanted something that was powerful, looked good, and most importantly Self-Hosted Infrastructure.
I wanted the power of Google Analytics, but for the physical world, running entirely on my own metal (Self-Hosted & Privacy-First).
So, I built Locus.
Turn RTSP streams into queryable data
Locus is an open-source project that brings professional-grade object counting and video analytics to EVERYONE. It's not just a script; it's a full-stack application designed to be easy to set up and even easier to use.
Under the hood, it’s combining the incredible speed of YOLO11 with ByteTrack for tracking. But you don't need to be an ML engineer to use it. You just spin it up, connect a camera, and start gathering data.
The "Cool" Stuff (Features)
🎨 Draw Your Own Zones
This is my favorite part. Every camera angle is unique, right? A reckless driver in a parking lot or a customer walking down a specific aisle—you can't capture that with a generic box. Locus lets you draw customizable polygon zones directly on the video feed. You can count exactly what happens in that specific corner of the frame.
🔒 Privacy by Default
I'm a huge believer in data privacy. With Locus, what happens on your network, stays on your network. There is no "cloud upload" feature. All processing happens locally on your CPU or GPU. Whether you're monitoring your home, a retail store, or a private facility, your footage never leaves your hardware.
⚡ Real-Time, For Real
We're using Next.js 15 for the dashboard, and it feels snappy. You get instant feedback on tracking, current occupancy counts, and dwell times. It’s mesmerizing to watch the analytics update live as objects move through the frame.
What Can You Actually Do With It?
I built this to be versatile. Here are a few improved ways people are using it:
- Retail Analytics: analyzing which store displays get the most attention (dwell time).
- Smart Home: counting how many cars pass by your driveway per hour.
- Safety: creating "danger zones" around machinery that trigger alerts if someone steps inside.
- Agriculture: setting up "virtual fences" to monitor livestock or detect predators on specific parts of a farm.
- Child Safety: detecting if a child enters a pool area when no adults are detected nearby.
- Crowd Management: triggering alerts when a zone exceeds a specific density of people.
- Logistics & Warehousing: tracking truck arrival/departure times and monitoring loading dock occupancy.
- Urban Planning: analyzing usage patterns of bike lanes versus pedestrian pathways to improve city infrastructure.
- Manufacturing: counting items on a conveyor belt to verify throughput or detect bottlenecks in real-time.
- Wildlife Monitoring: non-invasive tracking of animal crossings in nature reserves or backyards.
The Tech Stack
For my fellow developers, here is what makes it tick:
- Frontend: Next.js 15 (App Router), React 19, Tailwind CSS v4
- Backend: FastAPI (Python 3.10+)
- AI: YOLO11, ByteTrack, OpenCV
- Infrastructure: Fully Dockerized
Give It a Spin
Locus is open source and free. I'd love for you to try it out, break it, and tell me what you think.
Happy counting!