See everything. Understand everything. In real time.

Turn any camera into an intelligent eye. This app brings the power of cutting-edge computer vision AI directly to your cameras — detecting, counting, and tracking objects in a live video stream with stunning accuracy, all viewable from your browser in real time.

Whether you're monitoring a warehouse floor, counting vehicles on a highway, keeping an airport baggage belt running smoothly, or watching wildlife in the field — this is the only app you need.

Open your browser and see your camera feed — enhanced with real-time AI detections. Every object is identified, labelled, and highlighted with bounding boxes, right on top of the live video. The stream is ultra low-latency, so what you see is what's happening right now.

Choose from over 500 pre-trained AI models spanning 30+ detection architectures. From people and vehicles to animals, furniture, food, and industrial equipment — there's a model for virtually every scenario. Browse models by category, read descriptions of what each one is best at, and switch models on the fly.

Need to detect something unusual? With open-vocabulary models like Grounding DINO and YOLO-World, simply type what you're looking for — "red hard hat", "forklift", "unattended bag" — and the AI will find it. No retraining required.

Don't just detect — focus. Draw polygon zones directly on the live video to count objects only in the areas that matter to you. Set up counting lines to track how many objects cross a boundary and in which direction. Perfect for:

• Counting people entering or leaving a room
• Monitoring items on a conveyor belt
• Tracking vehicles passing through an intersection
• Watching inventory levels in a storage zone

Add as many camera streams as you want. Each stream gets its own model, its own settings, and its own detection zones. Manage them all from a clean, intuitive gallery view. USB cameras, network cameras (RTSP/HTTP), YouTube streams — they all work.

Every stream is fully configurable:

• Confidence threshold — control how certain the AI needs to be before reporting a detection.
• Class filtering — choose exactly which object types to detect and ignore the rest.
• SAHI sliced inference — enable tiled detection for spotting tiny objects in high-resolution footage.
• Object smoothing — reduce jitter with temporal averaging across frames.
• NMS tuning — eliminate duplicate detections with adjustable overlap thresholds.

All detection data — zone counts, line crossings, annotated snapshots — is automatically collected and published to a shared cloud dashboard. See trends over time, compare cameras, and get a bird's-eye view of what's happening across your entire camera fleet. Data from all devices in your swarm flows into one unified view.

This app runs entirely on your device — your video never leaves your network unless you choose to publish snapshots to the dashboard. AI inference happens locally on the NVIDIA GPU with hardware-accelerated TensorRT, delivering real-time performance even on compact Jetson devices.

For development and testing, the app also runs on standard laptops and desktops without a GPU.

Deploy this app on one device or a hundred. Each device in your IronFlock fleet runs independently with its own cameras and models, while all detection data flows into a shared cloud dashboard. Add new cameras and devices at any time — the system grows with your needs.

Install this app on your NVIDIA Jetson device through IronFlock, connect a camera, and open the web interface. You'll be watching AI-enhanced live video in minutes — no configuration files, no command lines, no complexity.

No camera? No problem. The app includes a built-in demo video so you can explore every feature right away.