The Road That Came With Proof
Vision AI · Roads · Mining · Infrastructure · Safety
Your cameras already see the site.
Photon understands what it means.
Photon is Cognecto's visual intelligence layer — connecting to your existing cameras to verify compaction coverage, detect cement spread anomalies, classify road defects, measure mine loading efficiency, and monitor site safety. All processed locally at the edge. Trained on 25,000 km of roads and 20+ mines across India, Australia, and Africa. Active in 7 days.
LIVE · CAM-07 (Roller Ch. 44+300) · Stream: DECODED · Compaction pass 11/14 verified · last sync: 2 secs ago
Road construction · Mining · Safety
25,000 km
Of road construction footage used to train Photon's visual intelligence models — the largest domain-specific dataset for infrastructure Vision AI in existence.
Mines across India, Australia & Africa
20+
Active mine deployments where Photon monitors loading efficiency, haul road operations, and site safety — generating the training data no generic Vision AI platform has.
Edge processing latency
150ms
From camera frame to verified event — processed entirely on-site. No cloud upload. No bandwidth cost. No lag between what the camera sees and what the dashboard shows.
What makes Photon different
Not a surveillance camera.
Not a surveillance camera.
A visual intelligence engine.
While worksite cameras record, Photon interprets—turning a passive video archive into active field intelligence.
When a roller compacts a road, Photon verifies lane coverage, chainage, and sequence. For cement spreading, it measures uniformity and flags deviations immediately. When haulers load under excavators, it estimates payload to alert under-loading in real time.
This specificity comes from training on 25,000 km of road construction footage and operational data from 20+ mines across three continents. Photon is built specifically for physical infrastructure.
Compaction pass lane coverage — verified against BOQ specification
Mining payload estimation — excavator-tipper cycle intelligence
Road defect classification — post-construction asset health monitoring
Site safety — PPE, zones, fire, and proximity across every shift
How it works
From camera frame
From camera frame
to structured field intelligence. In milliseconds.
Step 01 · Connect
Any camera. Any site.
Photon integrates with your existing IP cameras and NVRs via RTSP, RTMP, and ONVIF. No hardware replacement. No new infrastructure. If it streams, Photon processes it — whether it's a site security camera, a roller-mounted unit, or a fixed corridor monitor.
→
Step 02 · Analyse
Processed locally. Interpreted intelligently.
A ruggedized edge gateway on-site processes every frame using Cognecto's domain-trained models — compaction, cement spread, road defect, payload, and safety classifiers running simultaneously. No cloud upload. No bandwidth cost. No data leaves your network.
→
Step 03 · Alert
Zero-latency warning system.
The moment a safety violation is detected, Photon's alert engine fires within milliseconds — SMS, WhatsApp, and Cortex dashboard simultaneously. The event is saved to an immutable, cryptographically signed ledger for compliance auditing.
What Photon sees — beyond safety
Six use cases.
Request a demo →
Six use cases.
Trained on 25,000 km of roads
and 20+ mines across 3 continents.
Most Vision AI platforms stop at detecting people. Photon is trained on the specific visual signatures of infrastructure work — what a completed compaction pass looks like, what uneven cement spread looks like, what a loaded haul truck at 98% vs 60% capacity looks like. That specificity is the product of 25,000 km of road construction footage and real operational data from 20+ mines across India, Australia, and Africa.
Roads & Highways · Construction
Roller Lane Coverage Verification
Confirm actual coverage across all lanes rather than just movement. Missed strips, overlap deficiencies, and off-corridor passes are flagged in real time and linked to the BOQ before pouring the next layer.
Roads & Highways · Quality Control
Cement Spread Uniformity Detection
Identify thickness anomalies and coverage gaps in real time, geo-tagged and timestamped before the roller passes or the next layer covers them. Eliminates post-pour disputes.
Roads & Highways · Post-Construction
Road Surface Defect Classification
Transitions post-construction to scan for cracks, potholes, rutting, and edge raveling, geo-tagged to precise chainages. Defect data feeds Cortex to prevent high reactive repair costs.
Roads · Mining · Infrastructure
Construction Progress Visual Recording
AI-interpreted evidence of what was built, when, and by which assets. Logs layer completion against BOQ line items to flag incomplete work, replacing challengeable manual reports.
Mining · Production Analytics
Loading Site Efficiency & Payload Estimation
Monitor excavator-tipper cycles with sub-second precision. Estimates payloads, detects under-loading, measures bucket times, and tracks idle time to maximize daily output.
All Sites · All Shifts · All Conditions
24/7 Site Safety — PPE, Zones & Hazards
Enforce PPE and no-go zones around cranes and excavations. Proximity alerts warn workers near heavy machinery, and early fire detection triggers before heat sensors trip. Logs events in Cortex.
Under the hood
Built for the field.
Built for the field.
Not for the data centre.
Photon is engineered for remote, bandwidth-constrained infrastructure sites. All intelligence runs locally at the edge, sending only structured metadata rather than raw video to the cloud.
10+
Parallel streams per edge gateway — zero frame drops
100%
On-premise analysis — no video leaves your network
<150ms
Frame capture to alert dispatch latency
AWS
Fully integrated with AWS IoT Core and Cortex
Edge-first frame processing — Decodes and analyzes video frame-by-frame on-site. Continues monitoring during network outages, syncing logs on reconnect.
Privacy by design — No CCTV streams leave your local network. Photon only uploads anonymized event crops, timestamps, and statistics.
Encrypted tunnels — Zero open firewall ports or static IP needs. Full encryption between edge gateway and the Cortex dashboard.
Heavy-industry training — Custom models trained on thousands of hours of high-dust, rainy, and low-light scenes to minimize false alarms.
Stream & Camera Compatibility
Works with the cameras
Works with the cameras
you already have on site.
Photon is hardware-agnostic, connecting directly to your existing security, machine-mounted, or overhead cameras. Add instant intelligence without replacing any hardware.
Stream Protocols
RTSPRTMPONVIFWebRTCH.264H.265HTTP PushHLS
Camera Vendors
HikvisionDahuaAxis CommunicationsBosch SecurityMobotixHoneywellPanasonicGeneric IP CamerasOn-Site NVRsPTZ Cameras
In the field. On the record.
Photon, deployed.
Photon, deployed.
Real-world safety. Verified metrics.
50%
Near-miss reduction
3
Breaches in 48hrs
24/7
CCTV AI watch
"Bauxite mining involves heavy excavators, haul trucks, and on-foot workers in close proximity every shift. Photon monitors every crane swing radius and geofenced hazard zone. We identified three no-go zone breaches in the first 48 hours — events that would never have been caught under the previous manual regime."
50%
Near-miss reduction
3
Breaches in 48hrs
24/7
Automated site watch
Camalco · Cameroon Aluminium Company · Central Africa
Border Roads Organisation · Strategic Road Construction · India
Why PhotonWhy Photon sees
Why Photon sees
what no other system can.
Generic AI sees people; Photon is trained specifically on infrastructure to identify compaction passes, cement coverage, payloads, and defects.
01 / Cloud-Free, Edge-First AI
Edge-first. Cloud-free. Zero bandwidth cost.
All processing runs on local worksite edge nodes. With no streaming to the cloud, you eliminate bandwidth costs, latency, and compliance risks while uploading only metadata.
02 / Pre-Trained Industrial Models
Trained on infrastructure. Not the internet.
Unlike Vision AI trained on web images, Photon is trained on 25,000 km of road construction and 20+ mine sites in severe dust, rain, and low light.
03 / Proactive Prevention
Quality control happens before the next layer covers it.
Photon flags missed compaction lanes and cement anomalies in real time before they are buried, eliminating expensive post-construction repairs.
04 / Cortex-Integrated Audit Trails
Every observation becomes structured evidence.
Every detection is stored as a structured, geo-tagged event in Cortex. Searchable by chainage, shift, and asset, offering court-admissible records rather than video.
Part of a bigger picture
Boson hears.
Boson hears.
Photon sees.
Cortex decides.
Photon is the visual engine of Cognecto. Active on its own for safety, it connects with Boson and Cortex to form a comprehensive infrastructure operating system.
01 / Hear
Boson
IoT Sensing & Physics AI
Connects to any machine device, normalises telematics signals, and verifies field events with physics AI.
Explore Boson →
02 / See
Photon
Vision AI & Safety Monitoring
Computer vision running 24/7 across every site camera to detect PPE, intrusions, fire, and hazardous situations.
You are here
03 / Act
Cortex
Dashboard & Decision Layer
The operational layer — BOQ verification, shift planning, and role-based audit access from field to Ministry.
Explore Cortex →
Common questions
Everything you need
Everything you need
to evaluate Photon.
Don't see your question? Our team typically responds within 2 hours.
Request a demo →Do I need to buy new cameras to use Photon?
No. Photon is fully hardware-agnostic. If your existing cameras support RTSP, RTMP, or ONVIF protocols, Photon can ingest their feeds. We work seamlessly with Hikvision, Dahua, Axis, Bosch, and generic IP systems.
How does local edge-first AI processing work?
We install a ruggedized, pre-calibrated edge computer gateway on your site. This gateway decodes and analyses video streams locally. Only lightweight alerts and anonymized event metadata are sent to your dashboard — no heavy video upload, no bandwidth costs.
Is my site data secure? Does Photon record faces?
Your data is completely secure. Video processing is performed entirely locally — no streams ever leave your secure worksite network. Photon does not use facial recognition or track individual worker identities. It only classifies objects (helmets, vests, vehicles) to protect privacy while ensuring compliance.
What happens if our worksite internet connection drops?
Since Photon's AI runs on the local edge gateway, monitoring does not rely on internet connectivity. If the connection drops, Photon continues to scan, detect, and record events locally. Once connectivity is restored, it automatically syncs all historical logs and metadata to your dashboard.
How long does it take to deploy and go live?
On average, Photon is live across your site in 7 days. We configure your edge gateway pre-loaded with industrial models, plug it into your local camera switch, map your worksite geofences, and grant your safety team dashboard access.
How does Photon trigger alerts when a safety issue occurs?
When Photon detects a hazard, it generates a structured trigger within milliseconds — dispatching instant alerts via SMS, WhatsApp, and the Cortex dashboard simultaneously. The event is also saved as an immutable audit record for compliance tracking.
Visual intelligence. Active in 7 days.
Your cameras are already
Your cameras are already
collecting evidence.
Photon makes it count.
Connect to your existing cameras. Plug in the edge gateway. Configure your site models — compaction, spread, defect, payload, or safety. Live within 7 days. No hardware purchase. No custom training. Intelligence that starts on day one.
✓ Trained on 25,000 km of road construction
✓ 20+ mines across India, Australia & Africa
✓ Cloud-free edge processing
✓ Works with 95% of existing IP cameras
Frequently Asked Questions
Find answers to common questions about our platform and deployments.
Photon Vision is Cognecto's computer vision AI platform that processes live camera feeds from infrastructure sites. It uses deep learning models to detect safety violations, monitor material quality, verify construction progress, and track equipment utilization.
Photon Vision AI models are trained to identify PPE non-compliance (missing helmets, vests, harnesses), unauthorized personnel in restricted zones, proximity hazards near heavy equipment, and unsafe material handling practices from standard CCTV camera feeds.
Photon Vision works with any IP camera or CCTV system. It supports RTSP, ONVIF, and HTTP streaming protocols. Existing site cameras can be connected without additional hardware, or Cognecto can deploy ruggedized cameras for harsh environments.
Photon Vision achieves 95%+ accuracy on PPE detection and 92%+ on equipment classification in field conditions. Models are continuously retrained on site-specific data to improve accuracy for each deployment's unique environment and equipment types.