AI Camera with Kendryte K210——HUSKYLENS（AI Vision Sensor）

 1. Introduction to HUSKYLENS (AI Vision Sensor) 

   About HUSKYLENS

Built with the Kendryte K210 chip, HUSKYLENS (AI Vision Sensor) for DFROBOT is user friendly and easy on the wallet. This AI-powered camera positions all components on a small PCB (44.5 mm × 52 mm), with a 2 MP (or 5 MP) imaging sensor on the front and a 2" IPS display on the back.

Images shot by the imaging sensor (including faces, colors, and tags) are analyzed, recognized, and tracked by the Kendryte K210 on the printed circuit board. The UART/I2C delivers the output of identified and tracked objects (i.e., dimensions and coordinates).

HUSKYLENS is programmed with six algorithms, namely, face recognition, object recognition, color recognition, tag recognition, object tracking, and line tracking. When connected to the power supply, HUSKYLENS is immediately available for recognition and tracking shown by the display on the back. Fig. 1-3 illustrates face recognition (ID1) displays registered with HUSKYLENS.

 Picture 1-3. Facial recognition algorithms 
   External Connection

Its UART/I2C portal allows connection to a microcontroller, including micro:bit, Arduino, and Raspberry Pi. The dimensions and coordinates of a recognized object to be acquired on a real-time basis may help develop self-driving robotic vehicles and interactive applications in response to human faces and gestures.

HuskyLens can connect to Arduino, Raspberry Pi, LattePanda, or micro:bit

   HUSKYLENS Specifications

HUSKYLENS comes with a processor, imaging sensor, and display as shown below.

 1-1.HUSKYLENS Specification 

Processor	Kendryte K210
Image Sensor	SEN0305 HuskyLens：OV2640
Display	2.0-inch IPS screen with 320*240 resolution
Supply Voltage	3.3V～5.0V
Current Consumption	[email protected], [email protected] (face recognition mode; 80% backlight brightness; fill light off)
Connection Interface	UART/I2C
Dimension	52mm * 44.5mm
Built-in Algorithms	Object Tracking, Face Recognition, Object Recognition, Line Tracking, Color Recognition, Tag Recognition

   About Kendryte K210

The Kendryte K210 processor adopted by HUSKYLENS is a high-performance system-on-chip (SoC) developed by CANAAN CREATIVE for the AI and IoT markets.

  Image 1-4.Kendryte K210  

Kendryte K210 integrates various processors, including two FPUs powered by 64-bit RISC-V cores, neutral network processor (KPU), and audio processor (APU), as well as accelerators, including fast Fourier transform (FFT). The low-power Kendryte K210 can achieve high-performance image processing.

 2.Component Details 

The various components found within HUSKYLENS are described below.

   Imaging Sensor (Camera)

This imaging sensor (camera) is for shooting photographs. Modules for HUSKYLENS may vary from model to model. For instance, SEN0305 HUSKYLENS refers to the OV2640 (2-MP camera) and SEN0336 HUSKYLENS Pro refers to Ov5640 (5-MP camera). Upon purchase, the seal on the camera may be removed before use.
  Image 2-1.Image Sensor（Camera）
   Function Button

The function button is reserved for adjusting, changing, and confirming the settings menu. When the function button is pressed, the settings menu displays. You may select items by swiping left or right.

  Image 2-2.Function Button 

In addition to regular settings including communications protocols and display brightness, you may configure settings related to various algorithms (including face recognition, object recognition, color recognition, tag recognition, object tracking, and line tracking).

  Image 2-3.General Settings 
    Learning Button

The learning button is reserved for learning tasks. For instance, it is used to learn a human face and register such information with HUSKYLENS. In addition, the learning button allows shots from various angles to learn the human face. This enables face recognition from various angles.

  Image 2-4. Learning Button 
   UART/12C Portal

The 4-pin portal allows connection to an external microcontroller, including micro:bit, Arduino, and Raspberry Pi. The communications protocols include UART (9,600/115,200/1,000,000) or I2C. By default, the protocols can be fixed by auto recognition or manual configuration of settings.

  Image 2-5.UART/12C Portal  Form 2-1.UART Mode NumberLevelFunctionExplanation

1	T	TX	TX pin of HuskyLens
2	R	RX	RX pin of HuskyLens
3	-	GND	negative pole of power supply（0V）
4	+	VCC	positive pole of power supply（3.3V～5.0V）

 Form 2-2.I2C Mode NumberlevelFunctionExplanation

1	T	SDA	serial data line
2	R	SCL	serial clock line
3	-	GND	negative pole of power supply（0V）
4	+	VCC	positive pole of power supply（3.3V～5.0V）

   USB Portal

This Micro-B portal connects to the PC for power supply and firmware update of HUSKYLENS.
   Image 2-6. USB Portal    RGB LED

The LED lights up when the object is being recognized.

  Image 2-7.RGB LED 
   LED

Two (2) LEDs cast light on the object. Lighting adjustments, including on/off and brightness level, can be configured on the settings menu.
  Image 2-8.LED     Display

The display shows images shot by the imaging sensor. If the object is recognized, related details will display simultaneously. Furthermore, HUSKYLENS’ settings menu will appear.

  Image 2-9.Display  
  Image 2-10.Facial recognition  
   TF Card Slot

It is a slot for TF cards (MicroSD cards). The card is inserted following the arrow sign. Currently, its purpose remains unclear.

  Image 2-11.TF Card Slot  

 3.HUSKYLENS Documentation 

To learn more about HUSKYLENS, please see “HUSKYLENS_V1.0_SKU_SEN0305_SEN0336-DFROBOT”.