Spectro — AI Video Fire & Smoke Detection via Existing Surveillance Cameras

Design Solution · Fire Safety

Design Solution · Dream about it

AI-powered video analytics for fire and smoke detection using existing CCTV infrastructure.

Spectro applies machine-learning models to live surveillance video streams to detect visual signatures of fire and smoke, deployed via cloud processing without requiring new hardware installation. It addresses the retrofit cost barrier in fire detection by leveraging assets already present on most commercial and industrial sites. The system aims to achieve faster alert generation than traditional passive sensors (heat, ionisation) by recognising combustion signatures in real time.

Spectro's proposition is that machine-learning models applied to live CCTV feeds can detect fire and smoke signatures faster than passive thermal or ionisation sensors, and do so without new hardware, by treating existing camera networks as a detection layer. For retrofit sites with good camera coverage this is an attractive cost structure, and earlier visual detection of flame onset is a plausible advantage over sensors that respond to heat or particulate concentration. The evidence available, however, is entirely the vendor's own claims — no public deployments, no third-party accuracy benchmarks, no false-positive rate data, and no documented response latency under real fire conditions. Spectro is a very small company (2–10 staff) and the technical maturity and long-term support continuity are genuinely uncertain. The regulatory path is a structural constraint rather than a timing risk: building codes in most jurisdictions require EN- or UL-certified point detectors as life-safety devices, and video-based detection has not achieved equivalent status, meaning Spectro could not substitute for compliant hardware even if its accuracy were proven. Camera coverage gaps, network reliability dependencies, and cloud data-sovereignty concerns add further specification risks. This is a concept worth watching as the regulatory environment evolves, not a product ready for specification as primary fire detection.

Strengths

• No capital outlay for new sensor hardware at retrofit sites with existing CCTV
• Potential for earlier detection than passive thermal or chemical sensors by recognising visible flame/smoke onset
• Scalable cloud deployment model; no on-site infrastructure or maintenance burden
• Reuses existing camera network, reducing installation complexity and downtime

Considerations

• Entirely dependent on camera line-of-sight and lighting conditions; obscured, shadowed, or low-light areas become detection blind spots (high)
• Requires stable, low-latency network connectivity to cloud backend; offline or congested links degrade responsiveness (high)
• Camera resolution, frame rate, and optical properties of existing CCTV stock may be sub-optimal for accurate flame/smoke signature discrimination (moderate)
• Introduces ongoing cloud service dependency and recurring SaaS costs, shifting from capex-based sensor models to operational expense (moderate)

Risks

• No public deployments, pilot data, or third-party validation; detection accuracy, false-positive rate, and real-world latency remain undocumented (critical)
• Early-stage vendor (2–10 staff) with unverified technical maturity; viability, support continuity, and long-term roadmap are uncertain (critical)
• Regulatory acceptance unclear; building codes and fire safety standards may not yet recognise video-based detection as equivalent to EN/UL-certified sensors (high)
• ML model robustness to non-fire visual events (bright reflections, welding, steam, dust, fireworks) not publicly assessed; false-alarm risk uncalibrated (high)
• Cloud-hosted data processing raises privacy, cybersecurity, and data sovereignty concerns not addressed in available materials (moderate)
• No evidence of integration pathways to existing fire alarm systems, notification protocols, or emergency response workflows (moderate)

Performance

• Detection method: Cloud AI on existing CCTV streams
• New hardware required: None claimed
• Company size: 2–10 employees
• Grant ceiling (Serbia Innovation Fund): EUR 300k

Reality check

Cited only in a StartUs Insights aggregator article (curated list, not primary research). Company appears in Hello Tomorrow Deep Tech Pioneers (March 2025 cohort — announced in advance, not verified outcome) and claims Katapult24 and Serbia Innovation Fund pursuit (no confirmation found). Domain spectro-solutions.com returned no response at review. No case studies, regulatory approvals, independent test reports, or customer references published. No evidence of beta deployments or pilot data release. Accelerator selection alone does not confirm technical viability or market traction.

#video_analytics #machine_learning #early_warning #retrofit #cloud_processing #cctv_repurposing #fire_detection

Source