THERMAL AI FOR STRUCTURAL FIREFIGHTING
See through smoke.
Save more lives.
PyroSight is a helmet-mounted vision system that fuses thermal imaging, computer vision, and on-device AI — locating people, exits, and hazards when visibility drops to zero.
- Sensor-to-display latency
- 39 ms
- Sensor-to-display latency
- Continuous edge runtime
- 12 hrs
- Continuous edge runtime
EXIT A — 14 M
In-helmet display during a primary search — simulated feed.
Live Helmet Feed
What the Firefighter Actually Sees
A live reconstruction of the in-helmet display during a primary search: thermal fusion, victim detection, hazard prediction and AR navigation — with command receiving every frame. Inject a flashover or collapse and watch the system respond.
HELMET FEED STANDBY — SCENARIO TS-07
You are FF-014, entering a residential structure with near-zero visibility. PyroSight comes online seconds after entry.
FIREFIGHTER STATUS
FF-014 RIVERAHEART RATE
92 BPM
ENVIRONMENT
PYROSIGHT AI — ANALYSIS FEED
Model idle — awaiting helmet activation
ACTIVE HAZARDS — 0
No active hazards — continuous monitoring
COMMAND UPLINK
STANDBYE14-A RIVERAINTERIOR — SEARCH
E14-B CHENSTAGED — RIT
TRUCK 7VENTILATION — ROOF
AI Detection Showcase
Ten Detection Classes. Zero Guesswork.
PyroSight's vision models classify everything that matters in a burning structure — people, egress, hazards — each with confidence, distance and risk level.
Adult heat signature, crouched, low mobility — rescue priority one.
Live Command Center
Every Firefighter. Every Hazard. One Screen.
Incident command sees the full picture — live positions, heat mapping, fire spread prediction, rescue progress and team vitals — streamed from every helmet on scene.
INCIDENT #2261 — STRUCTURE FIRE — 4-STOREY RESIDENTIAL
68%
RESCUE
PROGRESS
FIRE SPREAD FORECAST
34%
TEAM STATUS — 4 UNITS DEPLOYED
CRITFlashover risk rising in sector B-2
INFOFF-02 entered north stairwell — floor 2
WARNO₂ below 60% — FF-03 rotation advised
INFOCivilian 2 handed to EMS triage
Capabilities
An AI Crew Member on Every Helmet
Sixteen systems working together in real time — hover any card to see how each one earns its place on the fireground.
Thermal Imaging AI
Sees heat, not light.
Radiometric thermal streams are denoised and enhanced on-device, resolving human signatures through zero-visibility smoke.
Edge AI Processing
No cloud. No lag.
A helmet-mounted accelerator runs every model locally — the system works deep inside concrete structures with zero connectivity.
Computer Vision
Purpose-built fire models.
Detection networks trained on thousands of live-burn scenarios classify what generic vision models have never seen.
Human Detection
Finds people in smoke.
Multi-frame fusion catches partial, occluded and prone body signatures — including children and unconscious victims.
Intelligent Navigation
Safest path, always.
A live path planner weighs heat, structure and debris to compute egress routes, re-planning multiple times per second.
Exit Recognition
Every way out, mapped.
Doors, windows and stairwells are catalogued on entry and continuously verified as conditions change.
Heat Mapping
Temperature everywhere.
Dense radiometric heat maps overlay the environment, flagging surfaces beyond PPE tolerance before you touch them.
Structural Risk Analysis
Predicts collapse.
Deflection, spalling and load-path cues are scored per structural element with live collapse probability estimates.
Multi-Sensor Fusion
Thermal + RGB + depth.
A fusion engine reconciles all sensor streams into a single coherent scene model — robust when any one sensor fails.
Real-Time Hazard Detection
Gas, arc, backdraft.
Continuous classification of combustion signatures, gas plumes and electrical arcing with instant crew-wide alerts.
Low-Latency Processing
Under 40ms glass-to-glass.
The full perception pipeline completes in under 40 milliseconds — overlays move with your head, not behind it.
Incident Recording
Every second, archived.
Synchronized video, telemetry and AI decisions are stored for training, review and post-incident investigation.
Voice Assistance
Hands stay free.
A noise-hardened voice interface reads out hazards and takes commands through SCBA masks and 110dB fireground noise.
Team Coordination
Shared situational map.
Every helmet contributes to one live map — teammates, victims and hazards visible to the whole crew.
Command Dashboard
Incident command view.
Chiefs monitor positions, vitals and fire behavior in real time, with AI-suggested tactical adjustments.
Rescue Analytics
Learn from every fire.
Post-incident analytics quantify search coverage, exposure time and decision latency to sharpen the next response.
Technology Architecture
From Photon to Decision
Watch data move through the stack — every stage lives on the fireground, not in a datacenter.
Helmet Cameras
Radiometric thermal + low-light RGB + depth, hardened to NFPA 1971
Edge AI Computer
40 TOPS accelerator riding on the helmet shell — fully offline
Computer Vision Models
Detection, segmentation and tracking tuned for fireground scenes
Thermal Analysis
Per-pixel temperature, hotspot tracking, flashover precursors
Navigation Engine
Hazard-weighted path planning, re-solved continuously
Augmented Reality Display
In-visor overlays rendered at 60fps with sub-frame latency
Incident Dashboard
Mesh-networked telemetry aggregated across the crew
Emergency Command Center
Full tactical picture for incident command and dispatch
Real-Time AI Pipeline
39 Milliseconds, Glass to Glass
One frame's journey through the perception stack — each stage lights up as the packet passes through.
Thermal Camera
2.1ms
RGB Camera
1.8ms
Sensor Fusion
3.4ms
AI Detection
11.2ms
Hazard Classification
4.7ms
Navigation Engine
6.3ms
Helmet Display
0.9ms
Command Dashboard
8.6ms
Product Showcase
The PyroSight Smart Helmet
Select a subsystem to open the hardware underneath — twelve of them packed into an NFPA-rated shell.
Thermal Camera
Radiometric LWIR sensor that resolves human heat signatures through zero-visibility smoke.
640×512 @ 60Hz · <40mK NETD
Technical Specifications
Built Like Equipment, Not Electronics
Every subsystem is specified against fireground reality — heat, water, impact, gloves, and zero connectivity.
SENSING
- Thermal imager
- 640 × 512 radiometric LWIR, 60 Hz, <40 mK NETD
- Optical camera
- 4K HDR, f/1.4, starlight-rated low-light pipeline
- Depth sensing
- Time-of-flight, 0.2–12 m range, ±2 cm
- Environmental
- 6-point shell thermocouple array, 0–600 °C
COMPUTE & DISPLAY
- Edge AI module
- 40 TOPS accelerator, 15 W, fully offline inference
- Perception latency
- 39 ms glass-to-glass, 60 fps overlay rendering
- AR display
- Waveguide optic, 4,000 nits, full color, SCBA-compatible
- Storage
- 512 GB AES-256 encrypted incident recorder
PLATFORM
- Weight
- 1.9 kg complete, balanced to NFPA helmet standard
- Runtime
- 12 h continuous; hot-swap battery in under 5 s
- Connectivity
- Self-healing crew mesh + LTE/5G backhaul fallback
- Environment
- IP67; shell rated to 260 °C / 15 min flashover exposure
Standards & Compliance
Designed and tested to meet:
- NFPA 1971 — Structural firefighting protective ensemble — shell platform
- NFPA 1802 — Two-way portable RF voice communications for emergency services
- MIL-STD-810H — Shock, vibration, thermal cycling and ingress test methods
- IP67 — Dust-tight; submersion to 1 m for 30 minutes
- AES-256 — End-to-end encryption for all telemetry, video and mesh traffic
- FCC Part 15/90 — Radio device authorization for mesh and backhaul radios
Certification programs in progress with accredited third-party labs. Current status matrix available to pilot departments.
Full specification sheet
PDF · rev 2.4 · includes integration guide
Mission Timeline
Four Minutes That Used to Take Ten
A reconstructed rescue, second by second — exactly as PyroSight logged it.
Firefighter Enters Building
T+00:00Engine 14 makes entry through the north stairwell. PyroSight begins mapping on the first step.
Smoke Fills Hallway
T+00:42Visibility collapses to under half a metre in forty seconds. Human vision is now useless.
AI Activates
T+00:44Thermal and RGB streams fuse into a live AR scene. The hallway reappears — as data.
Victim Detected
T+01:31A crouched heat signature resolves behind a bedroom door. Confidence: 96.4%.
Danger Zone Detected
T+01:58Flashover precursors flagged in the kitchen corridor. The zone is derated for routing.
Navigation Path Generated
T+02:04Safest route computed around the heat zone in 380 milliseconds and painted on the visor.
Victim Rescued
T+03:12Contact made, vitals stable. The carry-out route locks to every HUD on the crew.
Safe Exit Located
T+03:40West stairwell verified clear, painted green across the shared map.
Mission Complete
T+04:12All units out. Four minutes twelve seconds, door to door. Every decision logged.
Beyond the Fireground
Anywhere Vision Fails, PyroSight Works
The perception stack generalizes to any environment where seeing means surviving.
Industrial Rescue
Refinery and plant incidents where smoke, steam and toxic atmospheres overlap.
Mining
Sub-surface navigation and personnel location when dust kills every headlamp.
Military
Obscurant-penetrating vision for breach teams and combat rescue.
Search & Rescue
Wide-area thermal sweeps that find the missing in fog, night and debris.
Chemical Plants
Plume classification and safe-corridor routing during hazmat releases.
Oil & Gas
Flare, leak and hot-surface awareness across rigs and processing decks.
Warehouses
High-rack fire navigation where collapsing inventory rewrites the map.
Tunnel Rescue
Long-bore incidents with zero GPS — the mesh map is the only map.
Earthquake Response
Void-space victim detection inside pancaked structures.
Disaster Relief
Shared operating picture for multi-agency response in degraded zones.
Space Exploration
Helmet-native hazard perception for habitats where evacuation isn't an option.
Nuclear Facilities
Remote-monitored entries with dose-aware routing and full audit trails.
Autonomous Rescue Robotics
The same perception stack, mounted on machines that go where humans can't.
FAQ
What Departments Ask Us First
No. Every model — detection, thermal analysis, navigation — runs on the helmet's edge AI module. Connectivity only adds capability: when the crew mesh is up, helmets share one map and stream to command. When it isn't, each helmet is fully autonomous.
Get Started
Request a Pilot Program
We deploy with a limited number of departments each quarter. Tell us about your operation and our deployment team will reply within two business days.
EVERY PILOT INCLUDES
- Six-week supervised deployment — four helmets, no cost
- On-site fit, training and live-burn support from our field engineers
- Full command dashboard access for your incident commanders
- Joint data review and validation report with your command staff