CanMV K210 Mini Visual Recognition Module AI Accessory for Python Development Board Kit with 2.4-inch Display

CanMV K210 Mini Visual Recognition Module AI Accessory for Python Development Board Kit with 2.4-inch Display

Description:
- Built-in camera and touch screen display, frosted shell, smaller in size than a business card. Convenient deployment to various device scenarios.
- The CanMV K210 mini is developed by using for Python programming, making it easy to play various visual applications such as color recognition, object counting, barcode/QR code recognition, face detection, face recognition, number recognition, and self classification training learning without the need to master AI model algorithms. The reserved serial port/I2C interface of the module can be connected to the upper computer, STM32, ESP32, Arduino, Raspberry pie and other main control communication are used in combination.
- Provide over 100 self compiled online tutorials (continuously upgraded), related materials and code that are fully open source, and provide technical support. Facilitate user learning and secondary development, and achieve various AI visual recognition experiments and project applications.

Specification:
- Main control chip: K210
- Chip internal core: dual core RISC-V 64bit at 400MHz
- Neutral network processor: KPU (1TOPS)
- SRAM: 8MBytes
- Flash: 16MBytes
- SD card slot: Micro SD (up to 64G)
- Button: x2 (programmable button, reset button)
- LED: x2 (programmable LED light, power indicator)
- Serial port/I2C interface: XH - 1.25mm - 4P
- Sensor interface: XH - 1.25mm - 3P
- Power consumption: 5V at 200mA
- Display screen: 2.4-inch (resistance touch), 320 x 240 resolution
- Product size: 60.3 x 42.7 x 15mm; 63.9 x 46.3 x 15mm (with shell)
- Weight: 35g, 47g (with shell)
- Summary: The CanMV K210 mini is compact and easy to deploy to devices such as small cars and robots. It has a higher cost performance ratio and is suitable for pure visual application development. It can communicate with other MCUs through serial port. The appearance is a frosted shell and + acrylic protective plate.

Features:
- K210 high computing power chip: using K210 dual core RISC-V 64bit at 400MHz high performance processor, plus 1TOPS computing power KPU.
- High quality camera: use GC0328 camera for clearer imaging and better recognition.
- Integrated design of touch screen display: equipped with a 2.4-inch color screen (resistive touch), making it convenient to directly observe experimental results.
- Shell protection: frosted acrylic protective shell. Effectively protect the entire machine.
- Multi platform communication: by extending the serial port, visual recognition results can be sent to development boards with standard serial ports such as computer-upper computer, STM32, ESP32, for Arduino, and Raspberry Pi, achieving linkage.
- Supports multiple bracket extensions: supports official fixed brackets, angle adjustable brackets, and height + angle brackets, making it easy to assemble and match into devices such as small cars and robots.
- It is recommended to use for MicroPython to develop. For MicroPython is a simplified implementation of the Python 3 language. Users can easily use for Python programming to achieve bus control such as GPIO and serial port 12C, machine vision, as well as sensor and other peripheral module development.

Multiple Visual Recognition:
- Face detection: Face detection realizes the recognition of individual or multiple faces in an image, uses rectangular boxes to mark them, and returns the coordinates of the face position.
- Facial contour: After recognizing the face, the facial contour is drawn using 68 key points, and the result is marked with a rectangular box. At the same time, the position coordinates of the key points of the facial contour are returned.
- Facial features: Facial feature detection is used to detect some basic features of the face, including gender, whether the mouth is open, whether it is smiling, and whether it is wearing glasses. The result is marked with a rectangular box, and the facial feature position coordinates are returned.
- Facial recognition: Facial recognition is used for facial input and recognition of different personnel, such as company attendance machines, various facial recognition gates, etc. Implement facial input through code, mark the recognition result with a rectangular box, and return the facial position coordinates and the facial recognition result label.

Human Body Part Recognition:
- Human body detection: Identify the human body in the image, suitable for security monitoring and personnel count. The recognition result is marked with a rectangular box, while returning the facial position coordinates and the face recognition result label.
- Head detection: Identify the head of the human body in the image. The recognition result is marked with a rectangular box, while returning the facial position coordinates and the face recognition result label.
- Hand detection: Identify the hands of the human body in the image. The recognition result is marked with a rectangular box, while returning the facial position coordinates and the face recognition result label.

Color Recognition:
- Single color recognition: achieving single color recognition through preset color thresholds, such as recognizing a certain color of red, green, or blue. The recognition result is marked with a rectangular box and the relevant coordinates are returned.
- Multiple color recognition: Based on a single color recognition, improve the code to achieve simultaneous recognition of multiple colors through preset color thresholds. The recognition result is marked with a rectangular box, and the relevant coordinates and color results are returned.
- Object counting: Based on color recognition, modify the code to achieve object recognition of the same color by setting a preset color threshold. Thus achieving recognition of the number of objects. The recognition result is marked with a rectangular box, and the relevant coordinates and quantity results are returned.
- Car tracking: Car inspection is based on color recognition, where the camera acquisite a straight line image and calculates the deviation angle from the center to maintain the center position, thus achieving line inspection. The recognition result is marked with a rectangular box, while returning the relevant coordinates and results.

Code Recognition:
- Barcode recognition: Implement barcode recognition for multiple formats, use rectangular boxes to mark the recognition results, and return relevant coordinates and barcode corresponding values.
- QR code recognition: Realize the recognition of QR codes, mark the recognition results with rectangular boxes, and return relevant coordinates and corresponding information of the QR code.
- April Tag machine code recognition: Implement recognition of April Tag machine codes, mark the recognition results with rectangular boxes, and return relevant coordinates and machine code corresponding information.
- Number recognition: Implementing handwritten digit recognition based on the MNIST dataset model. The recognition result is marked with a rectangular box, and relevant coordinates and numerical information are returned.

Object Recognition:
- Identify objects in images based on the 20class model. The 20 class model refers to 20 pre trained YOLO2 models, including airplanes, bicycles, birds, boats, cars, cats, etc., which are loaded and recognized through the K210 visual module. The recognition result is marked with a rectangular box, while returning relevant coordinates and object information.

Self Classification Learning:
- By taking no less than 5 images of several objects and using K210's built-in KPU to train the features of each object, recognition and classification of custom objects can be achieved. The recognition result is marked with a rectangular box, while returning relevant coordinates and object category information.

K210 Chip Parameter:
- Internal core: RISC-V Dual Core 64bit, with FPU
- Main frequency: 400MHz (can overclock to 600MHz)
- SRAM: built-in 8M Bytes
- Image recognition: QVGA at 60fps/VGA at 30fps
- Voice recognition: microphone array (8 mics)
- Network model: support for YOLOv3/Mobilenetv2/TinyYOLOv2/facial recognition, etc.
- Deep learning structure: support for TensorFlow/Keras/Darknet/Caffe, etc.
- External: for FPIOA, UART, GPIO, SPI, I2C, I2S, TIMER
- Video processing: KPU, FPU, APU, FFT

Online tutorial:
- https://wiki.01studio.cc/

Package Included:
- 1 x CanMV K210 Mini Board
- 1 x Type-C Cable
- 2 x Adapter Cable (3Px1, 4Px1)