The detection stack that runs in GOTEAM field units is available on its own — a software SDK you embed in your own camera, your own vehicle kit, or your own command-and-control system. Camera-agnostic, runs at the edge, speaks SAPIENT. No GOTEAM hardware in the loop.
Same engine, same field-captured thermal corpus and the same SAPIENT output as the GOTEAM system — licensed as software instead of delivered as a unit. What follows is what you get, who it is for, the shapes it deploys in, and how to try it on your own footage.
// what you license
// detection engine
Thermal drone detector
A neural detector trained on 16-bit thermal imagery straight off uncooled LWIR cores — not on re-normalised 8-bit video. It takes the frames your sensor already produces and returns boxes with confidence, frame by frame. Trained on GOTEAM's own field-captured thermal corpus.
// tracking + geometry
Tracks and range estimate
Persistent multi-target tracks across frames with trajectory history, plus a per-target range estimate derived from box size and your camera's field of view. Output per track: bearing, elevation, estimated range, and a confidence history you can threshold yourself.
// false-alarm control
Clutter, cloud and birds
The parts that decide whether an operator trusts the system. Static-clutter learning suppresses persistently hot fixed objects — roofs, machinery, sun-warmed structures — while keeping a hovering target. A sky/cloud preprocessing stage holds contrast against broken cloud. Flight-behaviour and shape cues down-weight birds rather than deleting them.
// integration
Integration points
Frames in, tracks out. Detections and events are available as JSON over a local interface, as SAPIENT / BSI Flex 335 v2 messages for C2 interoperability, and as cropped imagery for operator review. You decide where they go: your software, a partner's C2, or a self-hosted GOTEAM hub.
// why a thermal leg
RF and EW layers find a drone by its radio link, so a fibre-optic-guided or fully autonomous drone gives them nothing to find — NATO's stated assessment is that EW counter-UAS systems are ineffective against fibre-optic drones, which combine high manoeuvrability, jamming immunity, and low visual and radar signature (JED). A thermal detector looks at the aircraft itself. If you are building a fused system, this is the leg that covers that gap.
// who licenses it
// 01
Camera & thermal-core OEMs
Add drone detection to a camera you already sell. The engine consumes your sensor's frames and returns tracks — the glass, the housing and the brand stay yours.
// 02
System integrators
You already deliver perimeter and site-security systems. Add a passive drone-detection layer on hardware you source yourself, integrated into the platform you hand over.
// 03
Primes & C2 owners
You own the operating picture and need a passive EO/IR leg inside it. Run the engine on your own node and take the tracks into your C2 over SAPIENT — no second console for the operator.
// 04
Platform & vehicle integrators
Mobile kits, masts, observation platforms and remote stations. An edge-class compute module and a thermal core are the whole hardware requirement.
// deployment shapes
Embedded in your product
The engine runs on your compute, inside your application. Detections never leave your system — nothing is sent to GOTEAM and no network is required.
A node into your C2
A self-contained detection node — your hardware or ours — that streams tracks into the C2 you already run, over SAPIENT / BSI Flex 335 v2.
Full self-hosted stack
Engine plus the GOTEAM hub, on your own servers: node map, live feeds, event review and multi-node roll-up, with no vendor cloud anywhere in the path.
Passive thermal
// sensing
Camera agnostic
// sensor input
Edge Jetson · Arc
// compute target
SAPIENT Flex 335 v2
// c2 output
Self-hosted
// no vendor cloud
Software only
// no goteam hardware
// how to get it
You do not have to take our word for any of the above, and you do not have to start with a sales call. Put your own thermal footage through the detector online, then pull its output into your own system with a scoped token. Commercials come last, once it has worked on your data.
// step 01
Evaluate online
Upload a thermal capture you recorded yourself and the detector runs on it — the same engine that runs in the field, not a demo build. You get back an annotated video, the per-frame detections, and range and bearing for each track. No signup and no login — pick a file and go. Nothing to install, no hardware to ship, and no call to book first.
Ask for a key and pull detections straight into your own system over HTTPS — no console, no export step. A key is read-only and scoped to named sensors, so it can never read one it was not granted. The versioned REST API covers sensors, detections with filters and cursor paging, a track's fix history, per-frame boxes as CSV, evidentiary media, a live SSE stream, and webhooks whose deliveries are HMAC-signed. Published as an OpenAPI 3.1 spec, with a zero-dependency Python client.
Licensing is scoped to hardware, volume and territory — per device or per site, embedded in your product or running as a node into your C2. We put a number on it once you have seen the detector work on your own data and we both know what the integration actually is. Software licensing is subject to end-user and export-control review, so expect those questions before a quote.
The fastest honest answer to "does this work on my scenes" is your own footage — send a thermal clip through FlyBox and the production detector processes it on real hardware, no signup and no login. Then take a scoped, read-only key and pull the same detections into your own system. Both start here; neither needs a sales call first.
Your own footageNo account neededAnnotated video backRead-only scoped keyOpenAPI 3.1 + Python client