Firefly CT36L/CT36B AI Smart Camera: $49 Industrial Edge AI Vision with 0.5 TOPS NPU

The Firefly CT36L/CT36B AI Smart Camera is an industrial-grade edge AI vision device powered by the Rockchip RV1106G2 processor. With a built-in 0.5 TOPS NPU, 3-megapixel sensor, and PoE support, it’s designed for smart surveillance, facial recognition, object detection, and edge AI applications.

RV1106G2 Chip: Edge AI at the Iron Price

At the heart of the CT36L lies the Rockchip RV1106G2, an ARM Cortex-A7 based vision processor running at 1.2GHz with an integrated 0.5 TOPS NPU. This isn’t a flagship AI chip — it’s a purpose-built solution for cost-effective edge inference.

Key processor specs:

  • CPU: ARM Cortex-A7 @ 1.2GHz with Neon and FPU
  • NPU: 0.5 TOPS, supports INT4/INT8/INT16 mixed operations
  • Framework support: TensorFlow, MXNet, PyTorch, Caffe, ONNX
  • VPU: 5M ISP @ 30fps, H.265/H.264 encoding
  • RAM: 128MB DDR3 (built-in)
  • Storage: 16MB SPI Flash

For makers and integrators, the 0.5 TOPS NPU is sufficient for running lightweight models like MobileNet-SSD, YOLO-tiny, or face detection — exactly what you’d want for a smart doorbell, parking sensor, or retail analytics device.

3-Megapixel Camera with Professional ISP

The CT36L features a 3-megapixel CMOS sensor (SC3336, 1/2.8″) paired with a 5-megapixel ISP capable of advanced image processing:

  • Resolution: 2304×1296 @ 30fps
  • Lens: 3.95mm focal length, F2.0 aperture
  • FOV: 98.3° wide-angle
  • HDR + WDR: High dynamic range for challenging lighting
  • Low-light performance: Multi-level noise reduction for clear night vision

The combination of wide FOV and HDR makes this camera suitable for outdoor surveillance where both coverage and dynamic range matter.

Industrial Design: IP67 Weatherproof

Unlike generic ESP32-camera modules, the CT36L is built for real-world deployment:

  • IP67 rated: Dust-tight and waterproof
  • Operating temperature: -20°C to 50°C
  • Humidity: 15%~90% RH (non-condensing)
  • Dimensions: 174mm × 57mm × 63mm
  • Power: DC12V or PoE

This is a camera that can be installed outdoors — on building exteriors, parking lots, or industrial facilities — without additional housing.

100M Ethernet with PoE: One Cable Does It All

The CT36L supports PoE (Power over Ethernet) through its 100M Ethernet port. A single Ethernet cable carries both power and data, dramatically simplifying installation. No power adapter needed at the camera site.

This is a key differentiator for permanent installations — run one cable, get power + network + video stream.

Linux OS: Full Software Freedom

Unlike many “AI cameras” that lock you into cloud APIs or proprietary apps, the CT36L runs Linux with full SDK access:

  • Build your own firmware with Buildroot or Buildroot SDK
  • Deploy custom AI models via RKNN Toolkit
  • Integrate with MQTT, RTSP, ONVIF — standard protocols
  • SSH access for remote management

Firefly provides complete documentation, SDK downloads, and a wiki with tutorials for:

  • Serial debug
  • Firmware upgrade (USB, SD card, MaskROM mode)
  • Linux kernel compilation
  • NPU model deployment
  • UART usage

Use Cases

What can you build with the CT36L?

Smart Surveillance

Local AI detection without cloud dependency. Detect people, vehicles, or custom objects entirely on-device.

Face Recognition Terminal

Build a face recognition door lock or attendance system. The NPU handles face detection/embedding locally.

Retail Analytics

Count customers, track dwell time, detect queue lengths — all processed at the edge.

Parking Sensor

Detect available parking spots, license plates, or vehicle types in real-time.

Industrial Quality Inspection

Run lightweight defect detection models for product quality control on the factory floor.

CT36L vs CT36B: What’s the Difference?

The Firefly store lists both CT36L and CT36B — they appear to be the same hardware with different naming conventions. Both use the RV1106G2 and share the same specifications.

Key Specifications Summary

Component Specification
Processor Rockchip RV1106G2, Cortex-A7 @ 1.2GHz
NPU 0.5 TOPS (INT4/INT8/INT16)
Image Sensor SC3336, 3MP (2304×1296)
Lens 3.95mm, F2.0, 98.3° FOV
Video H.265/H.264 @ 30fps
RAM 128MB DDR3
Storage 16MB SPI Flash
Network 100M Ethernet + PoE
OS Linux
Rating IP67 waterproof
Temp Range -20°C to 50°C

Resources

Conclusion

The Firefly CT36L/CT36B fills a gap in the edge AI market: it’s more capable than ESP32-camera modules but far more affordable than Jetson-based solutions. At a price point that allows volume deployment, it offers:

  • Local AI inference without cloud dependency
  • Industrial-grade hardware for outdoor use
  • PoE convenience for easy installation
  • Full Linux access for custom software

For makers building smart surveillance systems, for integrators deploying edge AI at scale, or for anyone who needs a capable vision module without the Jetson price tag — the CT36L deserves a look.

发表评论