To: k210 realizes face recognition (code interpretation attached)

be based on K210 Face recognition access control based on （ One ）
Enter the official website to get the face recognition source code https://maixpy.sipeed.com/zh/

---

One 、 Get the machine code
Click on maixhub, Go in and click obtain key_gen Machine code

Click to download this zip, After decompression, you will get a file named key_gen_v1.2 Of bin file , Will be bin Burn the file to the development board , Use kflash_gui Software , If you can't burn, please look for it in my last article .
https://blog.csdn.net/HuangChen666/article/details/113954656
（PS： If the following steps are not successful , Then it's best to erase the whole piece of firmware first , Then burn the firmware ）
Burn well key_gen_v1.2.bin after , Use any serial port terminal software , For example, before debugging 32 When using the serial port debugging assistant 、putty、MaixPy All terminals in the are OK , Here I use putty demonstration ：
Choose your own here com mouth （ There are usually two when the board is inserted com mouth , Select the previous ）, Then the baud rate is changed to 115200, Press the reset key of the board after opening , The terminal will output the machine code of your board , It's best to copy it and save it to txt In the document , Because it will be used later .

---

Two 、 Get face recognition model
Model get link
https://www.maixhub.com/modelInfo?modelId=14
Slide to the bottom and click the download button , If you don't have an account, you need to register .
Then paste your machine code and click “ Submit ”, You can get the face model zip.
Download to get a machine code as the name of kfpkg file , Use this file directly klash_gui Software burning to the development board .

---

3、 ... and 、 Run face recognition code
Face recognition code acquisition
link ：https://pan.baidu.com/s/1VU9ij7qnOGNLalbUCKZUiA
Extraction code ：sgdv

use MaixPy Connect the development board to run the code , So far, the basic face recognition function has been completed , The above code contains Chinese comments , It is convenient for everyone to understand the code , The above is also the source code provided by the boss on the official website , But this is just to let us know about the process of face recognition .
According to the code , It is mainly divided into the following processes ：
1、 Load various models
2、 function Face detection model , Find the face position in the picture and frame the face
3、 Convert the cut face image into kpu Received format
4、 function Face 5 Point key model , Get the left eye 、 Right eye 、 nose 、 Left corner of mouth 、 The position of the right corner of the mouth
5、 The original image and face image are Affine transformation , Transformation for Face image , Turn the face image into kpu Format
6、 Use Face 196 Witt eigenvalue model Calculate the... Of the face picture 196 Witt's eigenvalue , Calculate the final Facial features feature
Then compare the obtained face features with the previously saved face features to get a set of scores , Choose the largest score , And the score exceeds 85 branch （ You can set it yourself ） It is assumed that the person has been identified , And from... According to the corresponding subscript names Get the person's name in the list .

---

Four 、 summary
So far, the most basic face recognition function has been realized , But you will find that the functions provided by the source code are only given in advance in the code Mr.x full name , Obviously, it is impossible to realize the power-off storage of faces , If you want to realize the power-off storage of face feature values , We need to prepare one SD card , Save to flash You can also do so. , But memory is limited after all , It is recommended to save to SD card .
So let's share it here first , The next article will share how I realized the power-off storage of faces .

---------------------
author ： Morning Shao hc
source ：CSDN
original text ：https://blog.csdn.net/HuangChen666/article/details/113995079
Copyright notice ： This is the author's original article , Please attach a link to the blog post ！
Content analysis By：CSDN,CNBLOG One click reprint plugin for blog posts