TurboPi is a powerful, compact edge computing platform designed for AI inference, robotics, and computer vision applications. It integrates high-performance CPUs and NPUs with a robust I/O system to enable real-time processing in resource-constrained environments. Built to meet the growing demand for embedded AI at the edge, TurboPi bridges the gap between development boards like Raspberry Pi and industrial AI hardware.
TurboPi is an open source AI vision robot car designed for beginners. It is built on Mecanum wheel chassis, powered by Raspberry Pi 4B, loaded with 120-degree wide angle camera and uses OpenCV to process images. Equipped with RGB glowy ultrasonic sensor whose light can be adjusted, TurboPi is able to avoid obstacle smartly. In addition, it also supports variety of sensors to turn your creative AI ideas into reality.
Abundant tutorials and Python source codes are provided to help you advance quickly into robot technology, machine vision and artificial intelligence!

Core Components
GPIO, UART, I2C, SPI
Software Support
- Operating Systems: Ubuntu, Debian, Android, Yocto Linux
- AI Frameworks:
- TensorFlow Lite, ONNX Runtime, PyTorch (via conversion)
- RKNN Toolkit: Rockchip’s AI SDK for converting, optimizing, and running models on NPU
- Container Support: Docker, for deploying isolated AI/ML services
- Hardware Acceleration APIs: OpenCL, Vulkan, V4L2 (Video4Linux), GStreamer
Usage Examples:
Edge AI Inference for Object Detection
Use Case: Deploy real-time object detection models in industrial automation or smart retail.
Example Implementation:
- Deploy YOLOv5s model using RKNN Toolkit for conversion and optimization.
- Use a USB camera or MIPI CSI camera for video input.
- Real-time detection pipeline using OpenCV + GStreamer + NPU.
bashCopyEdit# Convert ONNX model to RKNN
rknn_toolkit_lite convert yolov5s.onnx yolov5s.rknn
- Application: Detect defects in manufacturing parts, count items in smart shelves.
Robot Vision and SLAM
Use Case: Enable mobile robots or drones to navigate using vision-based Simultaneous Localization and Mapping (SLAM).
Example Implementation:
- Integrate RTAB-Map or ORB-SLAM2 with RGB-D camera.
- Fuse IMU + vision inputs for enhanced accuracy.
- Use ROS 2 on TurboPi to handle sensor fusion and mapping.
- Application: Warehouse AGVs, autonomous delivery robots.
License Plate Recognition (ANPR)
Use Case: Build an Automatic Number Plate Recognition (ANPR) system for smart parking or traffic enforcement.
Example Implementation:
- Use a combination of YOLO (plate detection) and CRNN (character recognition).
- Capture frames from surveillance camera via RTSP.
- Run both models sequentially on NPU.
pythonCopyEdit# YOLOv5 plate detection + OCR on TurboPi
results = plate_model.detect(frame)
plate_text = ocr_model.infer(crop_plate(results))
- Application: Entry-exit automation in gated communities or smart toll booths.
Smart Inventory with Barcode + OCR
Use Case: Use AI at the edge to scan barcodes and read text on packages in logistics centers.
Example Implementation:
- Use OpenCV + ZBar for barcode recognition.
- Integrate Tesseract OCR or optimized CRNN model for label reading.
- Push results to MQTT broker or local SQLite DB.
- Application: Automate package intake in last-mile delivery centers.
Smart Surveillance with AI Analytics
Use Case: Run real-time video analytics for motion detection, intrusion alerts, and person recognition.
Example Implementation:
- Use OpenVINO-converted models or RKNN for face detection.
- Trigger alerts via HTTP, MQTT, or WebSocket on detected intrusion.
- Store clips locally and sync with cloud.
bashCopyEditffmpeg -i rtsp://camera -vcodec h264 -vf "draw_box on motion" -o output.mp4
- Application: Campus surveillance, home security, perimeter monitoring.
Agricultural AI – Plant Health Monitoring
Use Case: Detect diseases or pests in plants using edge AI to support precision agriculture.
Example Implementation:
- Train a CNN model on plant leaf images (e.g., EfficientNet).
- Deploy via RKNN Toolkit and analyze images from mounted cameras on drones or robots.
- Classify diseases like rust, blight, etc., with real-time alerting.
- Application: Crop disease detection in greenhouses or open fields.
Industrial IoT Gateway with AI
Use Case: Serve as an intelligent IoT gateway that pre-processes sensor data and runs AI for anomaly detection.
Example Implementation:
- Use serial/UART interface to collect data from PLCs/sensors.
- Run lightweight LSTM model to detect abnormal patterns.
- Send inference results via Modbus/TCP, MQTT, or HTTP to central SCADA systems.
- Application: Predictive maintenance in factories.
What makes it special?
Processor: Powered by Rockchip RK3588S – an 8-core ARM Cortex-A76/A55 CPU with integrated 6 TOPS NPU.
Memory: Up to 32GB LPDDR4/4x RAM.
Storage: eMMC module (16–128GB) + microSD + M.2 NVMe SSD support.
GPU: ARM Mali-G610 MP4 GPU – supports 8K video decoding and OpenGL/Vulkan.
AI Acceleration: Dedicated Neural Processing Unit (NPU) for AI/ML model inference.
I/O Interfaces:
USB 3.0, USB-C, HDMI 2.1, MIPI CSI/DSI
Gigabit Ethernet
Omnidirectional Movement, Flexible Field of View

- 360° Omnidirectional Movement
With 4 omnidirectional mecanum wheels, TurboPi can move 360°. Different movement modes (move forward, horizontally, diagonally and rotate ) and excellent performance make it bold to challenge various complicated routes.

- 2DOF Pan Tilt for 360° View
Equipped with 2 anti-blocking servos, pan tilt can rotate 130° vertically and 180° horizontally, which contributes to 360° camera field of view.
AI Vision Recognition, Target Tracking
TurboPi is powered by Raspberry Pi, takes OpenCV as image processing library, adopts Jupter Lab development tool, programmed in Python3 and armed with 2DOF camera pan tilt and DC TT motor. It is skilled in face tracking, color tracking, QR code recognition, gesture recognition, object recognition, license plate recognition, line following and many more AI games.
- Color Recognition and Tracking
Working with OpenCV, TurboPi can track specific color. After you select the color on the APP, it emits light of corresponding color and moves with the object of that color.
- Target Tracking, Different Tracking Modes
TurboPi tracks the target within its view with OpenCV. It supports two tracking modes. You can make pan tilt or pan tilt together with car body move with the target.

- Gesture Recognition, Man-robot Interaction
TurboPi cooperates with OpenCV to count fingers. Then interact with you based on the number of finger, such as honk horn, twist and change color of light.
- Face Detection and Tracking
With the help of OpenCV, TurboPi can track human face within its field of view, and move with it.
- QR Code Control
TurboPi can recognize and read QR code to perform related actions.

2 Driving Modes
- Visual Line Following
To achieve line following, it extracts ROI area with OpenCV first. Then remove noise to locate the line on binary image. Lastly, use PID algorithm to perform direction calibration.
- Following with 4-channel Line Follower
TurboPi uses 4 infrared probes on 4-channel line follower to detect and follow line.
Autonomous Driving
It combines 4-channel line follower and camera to achieve autonomous driving. Line follower is responsible for line following function, and camera is used to recognize traffic light and road signs.
APP Control, FPV Transmitted Image
Android and iOS APP are available for robot remote control. Via the APP, you can control the robot in real time and switch various AI games just by one tap.
- First Person View, HD Transmitted Image
TurboPi supports LAN and WiFi direct connection modes. After WiFi connection, the first person view will be transmitted to APP interface, which brings you more exciting and real robot control experience!

- Quickly Experience AI Games
WonderPi APP supports Android and iOS. Switch game modes easily and quickly to experience various AI games.
- Glowy Ultrasonic Sensor for Obstacle Avoidance
Loaded with glowy ultrasonic sensor with IIC interface, TurboPi can detect and avoid obstacle without occupying Raspberry Pi CPU.

Product Structure

- Hard Aluminium Alloy Chassis
It protects the core control board from shattering and shocking and can bear bigger load!
- High Performance Pan Tilt Servo
2DOF pan tilt is equipped with micro anti-blocking servo which has higher accuracy and longer service life.
- HD 120° Wide Angle Camera
Loaded with 120° wide-angle camera, TurboPi can look around 180°, and up and down 180°. Combined with mecanum chassis car, its visual range is up to 360°.
- 4-channel Line Follower
4-channel line follower with IIC interface enables TurboPi to detect black line without occupying Raspberry Pi CPU.

Powerful Control System
RaspberryPi 4B distinguishes itself in processor speed, multimedia performance, memory and connection. The combination of RaspberryPi 4B and RaspberryPi expansion board significantly enhances TurboPi AI performance!

Product Parameters

Packing List
