// AI SOFTWARE FOR DRONES & CAMERAS

Turn your drones and cameras into autonomous teammates.

AI-powered computer vision. Real-time thermal detection and object tracking, Mission Prompt, a control center and reports — ready software for your hardware, with no changes to the metal. For defense and civilian missions.

Request a demo integration See examples
// OUR MISSION Why we do this

We're building Bulgaria's sovereign visual capability. Autonomy at every level — from edge to cloud: a platform stack for defense and security that can see, think, adapt and collaborate in real time. A more advanced and more secure country — less load on operators, faster response, the decision always with the human.

Visual AI platform

VIDERE is our visual-AI platform for drones and cameras: real-time object detection and tracking on thermal and regular video, stream enrichment with metadata, and a single control center with Mission Prompt — the operator setting the task in plain language. Scalable and flexible, it runs on high-performance GPU servers — in the cloud or on your own infrastructure (on-premise), and on request on edge devices on board for autonomous operation. It integrates easily into third-party camera infrastructure over standard protocols and fine-tunes to your classes and sensors. The model isn't static: we collect data from real missions, fine-tune it and ship updates — the system continuously improves for your scenario, closing the loop from capture to a new model version. The solution is built for air, land and maritime platforms — for defense and civilian missions, with a human in the decision loop.

Real-time
Glass-to-glass latency
Thermal + RGB
Thermal and regular video
Drones & cameras
RTMP / RTSP / ONVIF · standard protocols
Human decides
Human-in-the-loop
On-premise
Your data stays with you
// 02 — HOW IT WORKS Stream → AI → operator → decision

Seamless integration: from video stream to decision.

A drone or camera sends video over standard RTMP/RTSP/ONVIF — it works with any video source, plug-and-play. The AI processes the stream in real time, enriches it with detections, tracks and metadata, and delivers it to the operator and the control center. Processing runs on a server or at the edge — the solution fits any hardware.

VIDERE · CONTROL CENTER · CH-01 LIVE
30 FPS GPU 41% TRACKS 04 ● AI ONLINE
  1. 01

    Video stream

    A drone or camera sends thermal or regular video over RTMP/RTSP/ONVIF — no proprietary hardware.

  2. 02

    Real-time AI

    Computer vision detects objects and thermal targets and tracks them across frames on the GPU.

  3. 03

    To the operator and control center

    Video with detections, tracks and metadata reaches the operator and the control center at the same time.

  4. 04

    Human decision

    The operator assesses and acts. The AI highlights and tracks — the decision stays with the human.

CAMERA-TO-SCREEN LATENCY
Real-time
Near real time — from the drone's shutter to the operator's screen.
// 03 — ONE PLATFORM What's in the system

Everything you need — in one system.

Every module works out of the box and adapts to your task — no need to assemble anything yourself.

01

Thermal detection

Detection on thermal video as the core of the platform: targets are visible at night, in fog and smoke, where an RGB camera is blind. This is what sets us apart.

02

Detection on regular video (RGB)

The same platform also works with regular color video: a detailed picture by day and in good visibility. Thermal and RGB complement each other — round-the-clock surveillance.

03

Object detection and tracking

Recognition, classification and counting of targets — people, equipment, vehicles, vessels — with stable tracking across frames.

04

Target coordinates and range

Every detected target is tied to ground coordinates and comes with a range estimate — the operator sees where the object is on the map and how far away it is.

05

Mission Prompt

Control detection in natural language: the operator states the task in words — 'watch people at the perimeter', 'count boats' — and the system rebuilds detection, filters and alerts for the mission. A capability unique to the market.

06

Control center

A navigator or control center receives real-time video from several sources and operators at once, and sees a single operational picture. Channel selection, switching between aircraft, one-tap recording.

07

Alerts and signals

The system raises the alarm itself when a target appears — the operator gets a signal instead of watching the screen non-stop. Alert rules are configured for the task, including via Mission Prompt.

08

Reports

An instant mission report right after it ends — a finished result, with no manual assembly or sifting through raw footage after the shift.

09

Any hardware

Connects to drones and cameras over standard protocols (RTMP/RTSP/ONVIF). Adaptation to your sensors, protocols and requirements on request.

// 04 — WHAT SETS US APART An innovative solution

Mission Prompt — intelligent mission control in natural language.

The operator states the task in words, and the system rebuilds detection logic for the mission: it filters objects, counts, sets priorities and cuts noise. Intelligent, innovative mission control — the key differentiator of the VIDERE system.

The AI rebuilds detection for the mission scenario in real time.

// 05 — APPLICATIONS Dual use — defense and civilian

One platform — dozens of scenarios.

VIDERE Aerial System is a visual-autonomy platform: real-time situational analysis, sensor data fusion and operator decision support. It works from a drone in the air and from fixed cameras — scaling from compact onboard devices to GPU servers and sitting on your existing camera infrastructure. For each task, detection is configured in words via Mission Prompt — across land, air and maritime scenarios.

UAV-CAM
● RECUAV-CAMAnalysing…
// 06 — DEPLOYMENT OPTIONS To fit your architecture

Three ways to deploy the VIDERE system.

Ready

Cloud / server (SaaS)

VIDERE Aerial System runs on our server or yours. The drone simply streams video — nothing to embed. The fastest start.

On request

Ground station (Local)

The software is deployed on a ground control station — with no load on the aircraft.

On request

On-board / on-camera (Edge)

VIDERE Aerial System right on the drone or on an edge device next to the camera — for limited-connectivity and autonomous scenarios.

We adapt quickly to your tasks, hardware, protocols and sensors. Need a feature that isn't there? We'll scope it and add it.

// 07 — SECURITY Data sovereignty

Your data never leaves you.

01

Encrypted channels

Communications and video transport — over encrypted channels and secure tunnels.

02

On-premise

Deployment on your servers — your data stays with you. Data sovereignty for government and defense customers.

03

Local deployment

A fully self-contained install in a protected loop, with no access to an external network.

04

Access control

Access management and licensing — who sees what, and when, in the system.

// 08 — PHILOSOPHY

To see is to understand in time.

VIDERE — Latin for 'to see'. We build not another algorithm, but vision that extends the human's where it fails: in the dark, in smoke, in a flow of events beyond a single gaze. The value of artificial intelligence is measured not by model complexity, but by the seconds it gives back to the human and the decisions it makes possible. The technology stays a tool — the final word is always the human's.

// 10 — WHY VIDERE How we differ from the market

Why VIDERE.

01

Thermal at the core

Thermal detection is the foundation of our system, not an option on top of regular video. Targets are visible at night, in fog and smoke, where RGB is blind.

02

Mission Prompt

Detection control in natural language. Intelligent, innovative mission control — the key differentiator of the VIDERE system.

03

Region

A local team on NATO's south-eastern flank. We understand the Balkans and Eastern Europe market — close by and in touch.

04

Affordability and flexibility

Affordable pricing and flexibility: we adapt quickly to your task, hardware and timeline — without long rollout cycles and bulky contracts.

05

Our own dataset

Our own data and continuous fine-tuning — the model evolves with the task instead of standing still.

06

Human-in-the-loop

A human makes the decision. Ethics and the NATO/EU standard — the AI highlights but does not act on its own.

// 11 — ABOUT Who's behind the system

A team that knows the field.

In defense and civilian security, trust is built on people. We're pilots and engineers who know both the mission in the field and the technology.

Presented at HEMUS 2026
OE

Olga Emelyanova

Founder & CEO

System and process architect, certified UAV pilot. Designs the platform's structure and workflows, connecting field tasks with the product and data.

IS

Igor Sorokin

Co-founder & CTO

AI engineer. Responsible for developing and training the AI models — detection on thermal video and its optimization.

VD

Veselin Dochkov

Co-founder · Strategic & Business Development

Leads strategy, external communications, talks with clients and partners, pilot projects and long-term contracts.

NT

Nikolay Trakiyski

AI Engineer · Agentic Systems

Builds the platform's agentic AI — the autonomous agent system behind Mission Prompt that turns the operator's intent into real-time detection logic, orchestrates detection and adapts the system to every scenario.

// 12 — FAQ

Frequently asked questions.

No — we don't ship a separate AI module to embed, but a complete, ready solution. VIDERE Aerial System is a finished computer-vision software platform for drones and cameras: real-time video ingest, thermal and RGB detection, object tracking, Mission Prompt, a control center, alerts and reports. You get a working turnkey system, not a library or API you have to build out and integrate yourself.
Many computer-vision systems look convincing in a demo but lose reliability in real field conditions. VIDERE is built around thermal AI detection — targets are distinguishable at night, in smoke and fog, where a regular RGB camera is blind. The key differentiator is Mission Prompt: the operator controls detection in natural language and the system rebuilds its processing logic for the task. On top of that — our own model, trained on our own dataset and fine-tuned to your scenarios and object classes. It's a complete system, not a standalone recognition feature.
No, detection is only part of the platform. VIDERE recognizes, classifies, counts and stably tracks targets (people, equipment, vehicles, vessels), ties them to ground coordinates and estimates range to the object. On top of that — rule-based alerts, control via Mission Prompt, a single control center for several sources and operators, and instant mission reports. A raw video stream becomes a ready operational picture for decision-making.
An in-house thermal AI detection system means an ML team, dataset collection and labeling, model training and validation, and GPU infrastructure: roughly 18–24 months and hundreds of thousands of euros a year before the first working result. VIDERE gives you a trained thermal AI and a ready platform around it today: deployment in weeks, adaptation to your drones, cameras, sensors and requirements, continuous model updates. It's many times cheaper and faster than building from scratch — while someone else's build drags on for years, you're already operating and delivering a finished system to your client.
No, that's not required. By default VIDERE runs on a server — in the cloud or on your own infrastructure (on-premise) — and the drone or camera sends the video stream over standard protocols. An on-board / on-camera option, where the AI runs directly on an onboard edge device next to the camera, is available on request — for limited-connectivity and autonomous scenarios.
VIDERE works with various drones and cameras that send video over the standard RTMP, RTSP and ONVIF protocols — with no equipment replacement. The platform integrates into your existing video-surveillance system and adapts to your sensors and protocols. Both thermal and regular RGB cameras are supported, on drones and on fixed posts.
Yes. The VIDERE model is fine-tuned to your object classes and scenarios — on our data and yours — so the system sees exactly what matters for your task. Integration of third-party ML models is possible on request.
Both. One VIDERE platform covers defense and civilian scenarios: reconnaissance and surveillance, border, site and perimeter security, patrol, situational awareness, event security, maritime security and rescue operations. Detection logic for a specific scenario is configured via Mission Prompt.
VIDERE can be deployed fully on your own servers (on-premise) — then video data and results stay inside your loop and are not sent to external networks. Encryption and secure communication channels are used, ensuring data sovereignty. For defense and government tasks, a self-contained install in a closed loop, with no external network access, is possible.
Yes. VIDERE's priority is thermal detection, the system's core competency: targets are distinguishable at night and in difficult conditions. The platform also works with regular color RGB video — by day and in good visibility. Thermal and RGB complement each other for round-the-clock surveillance. At the core — our own model and dataset with continuous fine-tuning.
VIDERE delivers practically real-time video — from the camera shutter to the operator's screen — so you can react to events during the mission instead of analyzing the recording after the flight. We keep reducing latency.
Mission Prompt is detection control in natural language. The operator states the task in plain words — for example, from 21:00 to 06:00 watch the perimeter, alert on vehicle or people movement, ignore animals — and VIDERE itself rebuilds detection, filters and alert rules for the mission. This is intelligent, innovative mission control and the key differentiator of the VIDERE system.
A human always makes the decision (human-in-the-loop). VIDERE highlights targets, tracks them and raises alerts, but does not act on its own — the final word stays with the operator. This approach aligns with the ethical standards and requirements of NATO and the EU for the use of artificial intelligence.
There are two licensing models, depending on deployment scale and data-confidentiality requirements. Subscription (SaaS): the platform is hosted in our cloud, billed by the number of simultaneously processed video streams. A license activates from a single stream; each stream (drone or camera) gets an individual license key for access control and accounting. This model suits a small fleet and civilian scenarios. On-premise license: for sensitive data and isolated environments the system is deployed fully on the customer's infrastructure — data never leaves the organization's perimeter, ensuring data sovereignty. The license is provided on an annual basis; for large deployments there are enterprise terms with a progressive per-stream discount and branding for the customer's organization. The exact configuration and licensing terms are defined for the customer's scenario.
The platform ships ready to operate: a source is connected over standard RTMP/RTSP/ONVIF protocols, with no hardware changes. Deployment and configuration on the customer's infrastructure take up to 14 business days, including adaptation to your cameras, sensors and protocols. If the model needs fine-tuning for domain classes or specific scenarios, a few weeks are added for data collection, fine-tuning and validation — a matter of weeks, not months or years.
// 13 — CONTACT Request a demo integration

Become part of the future of visual AI.

Leave a demo request — we'll show VIDERE in action on your scenario.

Required field.
Required field.
Enter a valid email.
Required field.
Request received.We'll get back to you within two business days.
Emailvesselin@skypatrol.bg
HeadquartersSofia, Bulgaria
EventHEMUS 2026 · exhibitor ↗