What Is SAI-HUB? Spectro AI’s On-Premises Automation Software Explained

The problem SAI-HUB is designed to solve

Enterprise inspection rarely ends at the camera. Teams need to plan repeatable routes, see live telemetry, run detection models, store evidence, find prior missions and deliver an alert or export to another system. Using disconnected tools for each step creates manual handoffs and unclear data paths.

SAI-HUB brings these functions into applications matched to the deployment. The common design goal is operational AI close to the sensor: local mission workflows, real-time detection and controlled data handling instead of a mandatory public-cloud pipeline.

• Plan and reuse inspection or patrol missions.
• Operate manually or automate supported routes and schedules.
• Run preloaded or custom computer-vision models.
• Review live video, detections, telemetry and stored evidence.
• Keep core processing and storage on-site with Brain-Box where required.

SAI-HUB RC, DD, VID and ROB compared

Each name identifies the operating surface. RC is for remote-controller drone workflows; DD is for DJI Dock operations; VID is for connected video streams; ROB is for mobile robots. This makes it possible to select only the application that fits the vehicle or sensor instead of forcing every deployment into one interface.

How SAI-HUB and Brain-Box work together

Brain-Box is the local hardware layer in the Spectro AI stack. It hosts SAI-HUB components, detection algorithms and storage close to drones, docks, cameras or robots. This architecture can keep operational video and detections inside the approved environment and reduce dependence on continuous internet access.

Two hardware formats address different deployments. Brain-Box M is intended for flexible field, controller, robot and video workflows. Brain-Box D is the dock-oriented form for DJI Dock 2 and Dock 3 installations. Exact sizing should reflect stream count, model workload, storage, enclosure and connectivity needs.

Which detection and data workflows are supported

Spectro AI provides computer-vision models for RGB and thermal inputs, including people, vehicles, animals, fire and smoke, safety gear, foreign object debris and solar panel hotspots. Available classes and deployment compatibility differ between on-premises and DJI on-board catalogues, so each use case should be checked against the intended aircraft and feed.

SAI-HUB applications can retain videos, detections and structured metadata locally. Depending on the configuration, outputs may include coordinates, object summaries, clips, logs and GIS-compatible files. A project should define the record schema and downstream integration before deployment.

• Preloaded models: ready starting points for common RGB and thermal targets.
• Custom training: a model developed around a specific asset, object or environment.
• Bring your own model: supported in selected workflows after format and performance validation.
• Local alerts: rules that surface a qualifying event to an operator or connected system.

How to choose the right SAI-HUB configuration

Choose from the operating environment backward. A pilot flying a Matrice or Mavic from a controller has different control and storage needs from an unattended Dock 3. A camera network may need several simultaneous streams and no flight planning at all. A mobile robot needs route and navigation logic.

The fastest way to qualify the fit is a workflow demonstration using your sensor, target and connectivity constraints. Confirm the exact hardware list, which functions stay local, supported model formats, alert path, log export and responsibilities for operational authorization.

1. Identify the vehicle, controller, dock or video protocol.
2. Define the target object or visible condition and required response time.
3. Decide whether inference belongs on-board or on Brain-Box.
4. Specify local storage, remote access and integration requirements.
5. Validate the full workflow with representative data before scaling.

Frequently asked questions