BRAINFRAME OS FOR DEVELOPERS
Real-Time Object Detection on Your RTSP Cameras, Running Locally
BrainFrame OS is an AI operating system for real-time processing of images and video streams. Connect the RTSP cameras you already have, run detection models on your own edge hardware, and get alerts when something happens in a virtual zone you draw. Bring your own detection model by packaging it as an OpenVisionCapsule.
How It Works
- Connect your RTSP cameras. Add each stream by its RTSP address. ONVIF cameras are supported.
- Choose the models. Install ready-made OpenVisionCapsules from the AI Marketplace, or package your own model as a capsule.
- Draw virtual zones and set alarms. Draw a region to count objects inside an area, or a line to count objects that cross it. Add an alarm with a condition, such as a person in the area, and the hours when it is active. Conditions can be based on a count or on a rate of change.
- React to the alerts. Read alerts through the REST API, or watch them in the BrainFrame client, where a stream with an active alert moves to the top of the grid.
What You Need
- IP cameras with RTSP streams.
- A computer running BrainFrame. The public release runs on x86-64 systems, and AI inference can run on the CPU, an integrated GPU, a discrete GPU or an NPU. A discrete GPU or NPU is optional. Inquire about a private release for ARM architectures. Video is processed on your own hardware.
What It Costs
BrainFrame Free costs nothing to download and includes two video streams. Resale, distribution and commercial use are allowed. BrainFrame Pro adds streams at $150 one time per additional stream, or $5 per additional stream each month. BrainFrame is licensed software. It is an open platform. The APIs, OpenVisionCapsules, many algorithms and BrainFrame Applications are open source.
See the pricing pageQuestions
- How do I run real-time object detection on RTSP camera streams locally?
- Install BrainFrame OS on a computer on the same network as your cameras and follow the steps above. The video is processed on your own hardware.
- How do I set up virtual zone tripwires and automated alerts on existing cameras?
- Draw a line zone on the video and add an alarm to it, with a condition and the hours when it is active. Alerts are available through the REST API.
- Can I deploy my own detection model on edge hardware?
- Yes. Package the model as an OpenVisionCapsule, the open source format BrainFrame uses for AI models. The developer documentation has tutorials for OpenVINO and TensorFlow object detectors. A model from another framework, such as a YOLO model, must first be in a form a capsule can run.