CamThink

What CamThink does

CamThink, a brand from Milesight, provides vision AI cameras and edge AI devices that enable developers, engineers, and makers to quickly create custom, production-ready AI models for real-world applications across industries. They simplify hardware customization and integration, streamlining the entire prototype-to-practice process. Combining expertise in hardware manufacturing with strong developer support, CamThink ensures ease of use and faster time-to-market for developers working on vision AI projects. Their focus is on providing multi-dimensional perception capabilities such as visual, audio, and environmental data, aiming to allow AI to understand the world more comprehensively and promote the popularization and development of edge intelligence technology.

Products

• NeoEyes NE503An open edge AI camera platform that combines high-resolution video input, local multi-model inference, containerized applications, and event-driven workflows for system integrators, algorithm developers, and OEMs. It turns 4K video input into on-device AI inference, application logic, and structured events, powered by Hailo-15H with 20 TOPS INT8 AI performance.
• NeoEyes NE101An ESP32-S3 sensing camera and low-power transmission node designed for battery-powered scenarios. It's a compact, ultra-low-power development board for scheduled or triggered visual records, built for remote monitoring workflows that need long battery life and flexible connectivity.
• NeoEyes NE301An STM32N6 edge AI camera with an on-device NPU for real-time inference. It's an event-driven edge AI camera designed for distributed sites where low maintenance and flexible deployment matter, offering rugged IP67 protection and industrial temperature resilience.
• NeoEdge NG4500An NVIDIA Jetson Orin AI box and industrial-grade edge AI box. It's a fanless Jetson AI box for multi-stream inference, offering up to 157 TOPS AI performance with Jetson Orin NX/Nano, JetPack 6.0+, and rich I/O.
• NE101 Wireless ModulesOptional Cat.1 LTE or Wi-Fi HaLow for long-range connectivity on NE101.
• OV5640 Camera ModulesCompact 5MP OV5640 camera module with selectable 60° or 120° FOV for embedded vision.
• Mounting Brackets for NeoEyes SeriesSecure, weather-resistant mounting brackets for CamThink NeoEyes Series (NE101/NE301).
• NE101 Dev KitA modular kit for learning, prototyping, and building IoT vision projects, allowing users to assemble their own NE101 Sensing Camera.
• NE101 Dev BoardAn open Type-C dev board for building IoT vision projects.
• NeoMind PlatformAn edge AI application platform for device management, real-time dashboards, rule engine, and AI Agent. It's LLM-powered edge automation for IoT, running fully offline.
• CinferA local AI inference API for NE301, enabling on-device object detection without cloud dependency.
• AI Tool StackAn open-source toolchain optimized for NE301 (STM32N6) for annotating, training YOLOv8, quantizing, and deploying a full edge AI pipeline.

Key capabilities

• Open, developer-friendly edge AI hardware
• Hardware customization and integration
• Local multi-model inference
• Containerized applications
• Event-driven workflows
• High-resolution video input (4K)
• On-device AI inference
• Structured AI events
• Sony IMX678 sensor
• Gen2 AI-ISP
• Motorized optical zoom
• Hailo-15H AI compute (20 TOPS INT8)
• OCI Containers support
• Custom models deployment
• Independent updates for applications
• Event Bus, Rule Logic, Structured Output
• MQTT, REST API, VMS integration, Local I/O for event sending
• Low-light color detail capture
• Balanced detail in harsh contrast
• Wide view to target detail adjustment
• Multi-model vision workloads (detection, OCR, pose, ReID, behavior analysis, visual search)
• Open hardware design, firmware, and developer tools
• Flexible AI deployment (device, gateway, platform, cloud)
• Rugged hardware options (battery, PoE, 4G, Wi-Fi, Wi-Fi HaLow, industrial I/O)
• Standard data outputs (MQTT, HTTP, APIs, SDKs)
• Custom engineering support (hardware config, firmware, interfaces, enclosures, OEM/ODM)
• Developer resources (documentation, GitHub, SDKs, tools)
• Low-power operation (NE101)
• Modular hardware (NE101)
• Wireless ready (NE101)
• Open SDK (NE101)
• On-trigger capture (NE101)
• Plug-and-Play Firmware
• Open Source Codebase
• IP67 protection
• Industrial temperature resilience (-20°C to +50°C)
• Flexible power (USB-C, battery, PoE)
• Flexible networking (Wi-Fi, LTE Cat.1, Ethernet)
• Compact and open hardware (NE301)
• Remote model update (OTA)
• Remote Wake, Instant Response (NE301)
• Configurable FOV (51° / 88° / 137° for NE301)
• JetPack 6.0+ support (NG4500)
• Dual Gigabit LAN, USB 3.x, DI/DO (NG4500)
• RS232, RS485, CAN (NG4500)
• HDMI @4K UHD, Audio I/O (NG4500)
• Wi-Fi 6, BT 5.0 (NG4500)
• Modular I/O Expansion Board (NG4500)
• Fanless design (NG4500)
• NVMe SSD ready (NG4500)
• 12–36V DC-in (NG4500)
• Ubuntu and Jetson DevFlow pre-installed (NG4500)
• CUDA, TensorRT and DeepStream support (NG4500)
• Event-driven low-power capture
• Scheduled field imaging
• On-device AI detection
• Remote configuration and OTA updates
• Structured event output
• Flexible field connectivity (LTE, Wi-Fi, Wi-Fi HaLow, PoE)

Use cases

• Creating custom, production-ready AI models
• Prototype-to-practice process streamlining
• Vision AI projects development
• Running vision AI where video starts
• Turning 4K video input into on-device AI inference, application logic, and structured events
• Deploying containerized applications for project-specific vision tasks
• Filtering model outputs, applying rules, and converting detections into structured AI events
• Sending structured events to trigger alarms, update platforms, and connect with field devices
• Building security, inspection, sensing, operations, and OEM systems
• Perimeter breach detection
• Restricted zone monitoring
• Intrusion detection
• Illegal dumping detection
• Vehicle loitering detection
• Vegetation/land condition recording
• Animal detection/species recognition
• Feeding area observation
• Fence perimeter status monitoring
• Watering point anomaly checking
• Behavior analysis
• PPE compliance monitoring
• Parking occupancy monitoring
• Program compliance monitoring
• Cooler status monitoring
• Checkout monitoring
• Construction site and temporary perimeter monitoring
• Hotel, warehouse, and facility security points monitoring
• Remote asset and unattended site monitoring
• Meter reading systems (OCR)
• Equipment status monitoring (utilities)
• Distributed inspection networks (smart cities)
• Wildlife monitoring systems
• Water-level and pollution alert workflows
• Visual sensing for remote field monitoring platforms
• Facility teams monitoring parking occupancy and space usage
• Industrial safety integrators building PPE compliance workflows
• Smart city operators analyzing public space activity
• Shelf availability and out-of-stock detection
• Planogram, display, and brand visibility monitoring
• Inventory status, freezer visibility, and store operation workflows
• Long-range visual sensing
• Non-contact water meter OCR
• Machine malfunction inspection
• Crop growth monitoring
• Livestock areas monitoring
• Remote and off-grid security monitoring
• Temporary perimeters, remote assets, vacant properties, and battery or solar-powered monitoring points
• Multi-point sites, construction areas, industrial yards, warehouse zones
• Fixed AI camera monitoring (PoE-powered)
• Predictive maintenance in smart manufacturing
• Real-time anomaly detection in smart manufacturing
• Quality inspection in smart manufacturing
• Warehousing and logistics optimization
• Smart retail applications
• Smart city applications
• Wildlife movement monitoring
• Water conditions monitoring
• Crop stress detection
• Smoke and waste detection
• Site anomalies detection
• Water level, land condition, crop status, habitat change, and site records
• Flood-prone areas, protected sites, reservoirs, facility boundaries, farms, and environmental risk monitoring

AI approach

CamThink provides edge AI cameras and devices that enable developers to create and deploy custom, production-ready AI models for real-world applications. Their hardware supports on-device AI inference, containerized applications, and event-driven workflows, reducing server dependency and enabling faster on-site decisions. They offer open-architecture hardware and open-source software, focusing on multi-dimensional perception capabilities including visual, audio, and environmental data. Their products integrate AI compute units like Hailo-15H and NVIDIA Jetson Orin/Nano, allowing for local multi-model inference and application deployment. They also provide an AI Tool Stack for training, quantizing, and deploying vision models directly to their devices.

Tech named: Hailo-15H, NVIDIA Jetson Orin, NVIDIA Jetson Nano, ESP32-S3, STM32N6, Sony IMX678, Gen2 AI-ISP, OCI Containers, MQTT, REST API, VMS integration, RTSP, YOLOv8, CUDA, TensorRT, DeepStream, JetPack, Docker, Arduino, ESP-IDF, Python SDK, Cinfer, AI Tool Stack

Industries served

• Computer Hardware Manufacturing
• Industrial IoT
• Security
• Infrastructure
• Agriculture
• Smart City
• Retail
• OEM
• Manufacturing
• Warehousing & Logistics
• Environmental Monitoring
• Utilities
• Construction
• Hospitality
• Research

What it says sets it apart

• Developer-focused brand
• Open, reliable, and deployment-ready edge AI hardware
• Simplifies hardware customization and integration
• Streamlines prototype-to-practice process
• Leverages Milesight’s hardware expertise and global supply capabilities
• Ensures ease of use and faster time-to-market
• Ability to run AI models, application logic, and event workflows directly on the camera
• Built beyond traditional IP cameras
• AI-ready visual input for real-world scenes (low-light, harsh contrast, variable working distances)
• Runs multi-model AI on the camera (detection, OCR, pose, ReID, behavior analysis, visual search)
• Complete vision applications can be packaged and run at the camera as OCI applications
• Open by design (hardware, firmware, developer tools)
• Flexible AI deployment options (device, gateway, customer platform, cloud)
• Ready for real sites with diverse power and connectivity options
• Built for existing stacks with standard data outputs
• Custom engineering support available
• Comprehensive developer resources (Wiki, GitHub, Hackster, Discord)
• Modular hardware with swappable components
• Open source code for customization
• Rugged IP67 protection and industrial temperature resilience
• Flexible power and networking to adapt to real-world infrastructure constraints
• Remote wake and instant response capabilities for power efficiency
• Seamless AI deployment with local or OTA updates, staged rollout, and rollback
• Dedicated AI Tool Stack for training, quantizing, and deploying models directly to devices
• NeoMind platform for offline, LLM-powered edge automation
• Supports various deployment patterns for remote field monitoring (standalone, gateway-based, hybrid)
• Strong industrial foundation from Milesight (manufacturing standards, quality control, global infrastructure, long-term availability, partner ecosystem)