Device side
RTSP, RTMP, GB28181, USB and other mainstream video sources are unified at ingestion.
FRAMEFOCUS / Safety Production Video Early Warning
Turn on-site video into a real-time safety warning system
Designed for mining, hazardous chemicals, metal smelting and construction sites, Zhiyan Origin combines device-side, edge-side and cloud-side deployment to convert video streams into actionable alerts, warnings and management workflows.
24/7
continuous edge monitoring
People · Assets · Objects · Environment
core risk coverage
50%
shorter deployment cycle
Deployment model
Edge side
Edge computing boxes or inference servers perform real-time analysis close to the scene.
Cloud side
Node management, alert aggregation, message forwarding and centralized visibility.
Industry context
High-risk operations need earlier smarter warnings
In mining, chemicals, smelting and construction, hidden risks are dense while pre-event and in-process warning mechanisms are often weak. At the same time, monitoring systems and camera vendors are fragmented, making unified management difficult.
As AI, IoT and intelligent video analytics mature, policy and enterprise demand are converging around smarter safety supervision. The need is no longer just to watch the site, but to understand it and trigger credible alerts before incidents escalate.
Architecture
From video ingestion to event response built as one loop
The solution balances real-time local inference, centralized operations and long-term extensibility through a three-layer deployment model.
Device side
Ingest existing video infrastructure
Compatible with mainstream protocols and monitoring equipment, enabling rapid onboarding of current camera assets.
Edge side
Analyze risk where it happens
Deploy edge boxes or inference servers based on stream count and operating constraints to reduce transfer cost and accelerate warnings.
Cloud side
Manage nodes and messages centrally
Aggregate alerts from multiple edge nodes and provide preview, forwarding, traceability and unified operational control.
Software platform
A video intelligence platform built for risk-heavy operational environments
Device management
Onboard cameras with one click, monitor online status in real time, and detect occlusion or displacement that could degrade alert quality.
Application management
Schedule start and stop, batch operations, batch device replacement and bulk rule updates to improve operating efficiency.
AI analytics
Preloaded standard detection models cover analysis for people, machines, objects and environments.
Alerts and monitoring
Push warning events as snapshots or short clips to other systems or mobile devices, with dashboards for live preview and review.
Scenarios
Build durable video supervision around people assets objects and environment
Operational compliance
Detect fatigue, smoking, phone usage, vehicle misparking and other non-compliant actions.
Protective gear compliance
Check helmets, uniforms, gloves and face protection to enforce safety dress requirements.
Environmental monitoring
Warn on open flames, smoke and water accumulation in operating environments.
Dangerous behavior alerts
Identify falls, restricted-area entry, climbing, crowd anomalies and unsafe proximity to suspended loads.
Equipment status recognition
Track working status, health conditions and instrument readings to support maintenance decisions.
Access and zone management
Count people, recognize faces, record vehicle information and summarize movement in sensitive areas.
Example use case
Move from manual inspection to 24-hour intelligent monitoring
The deck includes a case on detecting illegal wastewater dumping near street-side restaurants: edge computing monitors continuously, reduces city management labor and sends violation location plus animated evidence back to the operations center.
Hardware
Choose low-power edge deployment or high-throughput data room deployment
Edge Computing Box GNR-FF1
Built for desktop deployment and industrial environments
CPU plus AI accelerator architecture, zero-copy memory sharing, fanless thermal design and dust-proof, moisture-proof stability for warehouses, checkpoints, factories and remote locations.
Inference Server GNRDeepFFT4
Built for rack deployment and rapid expansion
With high-performance acceleration cards, a single unit can handle more than one hundred real-time streams and scale horizontally as workloads grow.
Advantages
Faster rollout lower transfer cost and more reliable on-site warning
Edge intelligence
Real-time analysis on site lowers transmission and low-value storage costs while improving alert timeliness.
Hardware-software optimized inference
Models, platform and hardware are tuned together to maximize throughput and deployment stability.
Plug in and go live
Preloaded warning capabilities make deployment faster than traditional custom on-site model delivery workflows.
Commercial model
Standardized products plus custom delivery for different project phases
Integrated hardware and software
- Edge Computing BoxCNY 36,000
- Inference ServerCNY 180,000
Includes platform software and standard models. FF1 supports up to 8 streams / 2 standard models; the inference server supports up to 100 streams / 5 standard models.
Platform and models
- Platform softwareCNY 45,000 / node
- Algorithm modelsCNY 2,700 / stream
The platform supports x86 + NV and Huawei Kunpeng + Ascend architectures. Standard model lists are available on request.
Custom development
- Standard model optimizationCNY 900 / 1,000 images
- Custom model developmentQuoted per case
- Platform customizationQuoted per case
Suitable for long-tail scenarios, third-party system integration and executive dashboard projects.
About
Building AI agents that enterprises can trust as operational partners
Zhiyan Origin is a frontier technology company focused on intelligent systems. We believe AI agents will become the core of future production and management by combining computer vision and natural language processing with observation, judgment and execution.
Our core product lines include behavior recognition, safety production, product quality inspection and intelligent robotics. From production lines to monitoring rooms, these systems help enterprises handle monitoring, analysis and decision support at scale.
FrameFocus
An AI-driven visual platform focused on behavior recognition and safety management, designed to observe environments precisely and trigger alerts or intervention when needed.
InstaForma Robot AI-Agent Platform
An AI-agent platform for different classes of robots, enabling autonomous perception, understanding and execution in logistics, intelligent service and other operating scenarios.
Contact
Need the full proposal model list or a deployment assessment?
We can map stream volume, operational conditions, deployment location and warning goals to the right mix of edge boxes, inference servers and platform capabilities.
Typical inquiries
- Safety video early warning solution deck
- Standard model list and customization scope
- Integrated deployment and third-party system integration