8.4 Face Recognition

Preparation for development

  • Complete the device link
  • Complete firmware burning
  • Complete software environment

Burn specified firmware

We use the MaixPy IDE to run this routine. First, we need to download the model we need from the MaixHub, which need us to provide the machine code, so we download the bin file that can get the machine code, then burn it token_gen bin file address

Get Machine code

Use kfalsh_gui to burn ken_gen.bin file,click it to download.

View the machine code in the serial output

After the firmware is burned, connect the computer, open the serial assistant, then open the serial port.

Download model

With the machine code we can start to download the corresponding model files.

Model download address

model download address

List of files to extract :

The blue part is the model file, and the yellow part is the bin file of MaixPy.This is the compact version, and the size is relatively small, just more than 600 KB. Inside the json file is the configuration, which is about where these files should be downloaded to Flash and whether they need to be checked.

    "version": "0.1.0",
    "files": [
            "address": 0,
            "bin": "maixpy_face_ide.bin",
            "sha256Prefix": true
            "address": ‭5242880‬,
            "bin": "FD_a6e91e13a0de48bafec324646d070358.smodel",
            "sha256Prefix": false
            "address": ‭6291456‬,
            "bin": "KP_chwise_a6e91e13a0de48bafec324646d070358.smodel",
            "sha256Prefix": false
            "address": ‭7340032‬,
            "bin": "FE_mbv1_0.5_a6e91e13a0de48bafec324646d070358.smodel",
            "sha256Prefix": false

Burn model files

Download the Kkfpkg file to our development board using kfalsh_gui and run it


Run example code

code address

import sensor,image,lcd  # import 相关库
import KPU as kpu
import time
from Maix import FPIOA,GPIO
task_fd = kpu.load(0x200000) # 从flash 0x200000 加载人脸检测模型
task_ld = kpu.load(0x300000) # 从flash 0x300000 加载人脸五点关键点检测模型
task_fe = kpu.load(0x400000) # 从flash 0x400000 加载人脸196维特征值模型
clock = time.clock()  # 初始化系统时钟,计算帧率
key_pin=16 # 设置按键引脚 FPIO16
fpioa = FPIOA()
key_pressed=0 # 初始化按键引脚 分配GPIO7 到 FPIO16
def check_key(): # 按键检测函数,用于在循环中检测按键是否按下,下降沿有效
    global last_key_state
    global key_pressed 
    if last_key_state == 1 and val == 0:
    last_key_state = val

lcd.init() # 初始化lcd
sensor.reset() #初始化sensor 摄像头
sensor.set_hmirror(1) #设置摄像头镜像
sensor.set_vflip(1)   #设置摄像头翻转
sensor.run(1) #使能摄像头
anchor = (1.889, 2.5245, 2.9465, 3.94056, 3.99987, 5.3658, 5.155437, 6.92275, 6.718375, 9.01025) #anchor for face detect 用于人脸检测的Anchor
dst_point = [(44,59),(84,59),(64,82),(47,105),(81,105)] #standard face key point position 标准正脸的5关键点坐标 分别为 左眼 右眼 鼻子 左嘴角 右嘴角
a = kpu.init_yolo2(task_fd, 0.5, 0.3, 5, anchor) #初始化人脸检测模型
img_lcd=image.Image() # 设置显示buf
img_face=image.Image(size=(128,128)) #设置 128 * 128 人脸图片buf
a=img_face.pix_to_ai() # 将图片转为kpu接受的格式
record_ftr=[] #空列表 用于存储当前196维特征
record_ftrs=[] #空列表 用于存储按键记录下人脸特征, 可以将特征以txt等文件形式保存到sd卡后,读取到此列表,即可实现人脸断电存储。
names = ['Mr.1', 'Mr.2', 'Mr.3', 'Mr.4', 'Mr.5', 'Mr.6', 'Mr.7', 'Mr.8', 'Mr.9' , 'Mr.10'] # 人名标签,与上面列表特征值一一对应。
while(1): # 主循环
    check_key() #按键检测
    img = sensor.snapshot() #从摄像头获取一张图片
    clock.tick() #记录时刻,用于计算帧率
    code = kpu.run_yolo2(task_fd, img) # 运行人脸检测模型,获取人脸坐标位置
    if code: # 如果检测到人脸
        for i in code: # 迭代坐标框
            # Cut face and resize to 128x128
            a = img.draw_rectangle(i.rect()) # 在屏幕显示人脸方框
            face_cut=img.cut(i.x(),i.y(),i.w(),i.h()) # 裁剪人脸部分图片到 face_cut
            face_cut_128=face_cut.resize(128,128) # 将裁出的人脸图片 缩放到128 * 128像素
            a=face_cut_128.pix_to_ai() # 将猜出图片转换为kpu接受的格式
            #a = img.draw_image(face_cut_128, (0,0))
            # Landmark for face 5 points
            fmap = kpu.forward(task_ld, face_cut_128) # 运行人脸5点关键点检测模型
            plist=fmap[:] # 获取关键点预测结果
            le=(i.x()+int(plist[0]*i.w() - 10), i.y()+int(plist[1]*i.h())) # 计算左眼位置, 这里在w方向-10 用来补偿模型转换带来的精度损失
            re=(i.x()+int(plist[2]*i.w()), i.y()+int(plist[3]*i.h())) # 计算右眼位置
            nose=(i.x()+int(plist[4]*i.w()), i.y()+int(plist[5]*i.h())) #计算鼻子位置
            lm=(i.x()+int(plist[6]*i.w()), i.y()+int(plist[7]*i.h())) #计算左嘴角位置
            rm=(i.x()+int(plist[8]*i.w()), i.y()+int(plist[9]*i.h())) #右嘴角位置
            a = img.draw_circle(le[0], le[1], 4)
            a = img.draw_circle(re[0], re[1], 4)
            a = img.draw_circle(nose[0], nose[1], 4)
            a = img.draw_circle(lm[0], lm[1], 4)
            a = img.draw_circle(rm[0], rm[1], 4) # 在相应位置处画小圆圈
            # align face to standard position
            src_point = [le, re, nose, lm, rm] # 图片中 5 坐标的位置
            T=image.get_affine_transform(src_point, dst_point) # 根据获得的5点坐标与标准正脸坐标获取仿射变换矩阵
            a=image.warp_affine_ai(img, img_face, T) #对原始图片人脸图片进行仿射变换,变换为正脸图像
            a=img_face.ai_to_pix() # 将正脸图像转为kpu格式
            #a = img.draw_image(img_face, (128,0))
            del(face_cut_128) # 释放裁剪人脸部分图片
            # calculate face feature vector
            fmap = kpu.forward(task_fe, img_face) # 计算正脸图片的196维特征值
            feature=kpu.face_encode(fmap[:]) #获取计算结果
            reg_flag = False
            scores = [] # 存储特征比对分数
            for j in range(len(record_ftrs)): #迭代已存特征值
                score = kpu.face_compare(record_ftrs[j], feature) #计算当前人脸特征值与已存特征值的分数
                scores.append(score) #添加分数总表
            max_score = 0
            index = 0
            for k in range(len(scores)): #迭代所有比对分数,找到最大分数和索引值
                if max_score < scores[k]:
                    max_score = scores[k]
                    index = k
            if max_score > 85: # 如果最大分数大于85, 可以被认定为同一个人
                a = img.draw_string(i.x(),i.y(), ("%s :%2.1f" % (names[index], max_score)), color=(0,255,0),scale=2) # 显示人名 与 分数
                a = img.draw_string(i.x(),i.y(), ("X :%2.1f" % (max_score)), color=(255,0,0),scale=2) #显示未知 与 分数
            if key_pressed == 1: #如果检测到按键
                key_pressed = 0 #重置按键状态
                record_ftr = feature 
                record_ftrs.append(record_ftr) #将当前特征添加到已知特征列表
    fps =clock.fps() #计算帧率
    print("%2.1f fps"%fps) #打印帧率
    a = lcd.display(img) #刷屏显示

#a = kpu.deinit(task_fe)
#a = kpu.deinit(task_ld)
#a = kpu.deinit(task_fd)

Run the program with MaixPy IDE

Program running as shown in figure


Press BUTTON A to record the face. After the face is recorded, the name will be assigned in order and displayed when the face is recognized.

results matching ""

    No results matching ""