Skip to main content

8. Mask Recognition Experiment

wmnologoAbout 2 min

Routine Code

import sensor, image, time, lcd
from maix import KPU

lcd.init()
sensor.reset()
sensor.set_pixformat(sensor.RGB565)
sensor.set_framesize(sensor.QVGA)
sensor.skip_frames(time = 1000)
clock = time.clock()

od_img = image.Image(size=(320,256), copy_to_fb=False)

anchor = (0.156250, 0.222548, 0.361328, 0.489583, 0.781250, 0.983133, 1.621094, 1.964286, 3.574219, 3.94000)
kpu = KPU()
print("ready load model")
kpu.load_kmodel("/sd/KPU/face_mask_detect/detect_5.kmodel")
kpu.init_yolo2(anchor, anchor_num=5, img_w=320, img_h=240, net_w=320 , net_h=256 ,layer_w=10 ,layer_h=8, threshold=0.7, nms_value=0.4, classes=2)

while True:
    clock.tick()
    img = sensor.snapshot()
    od_img.draw_image(img, 0,0)
    od_img.pix_to_ai()
    kpu.run_with_output(od_img)
    dect = kpu.regionlayer_yolo2()
    fps = clock.fps()
    if len(dect) > 0:
        print("dect:", dect)
        for l in dect :
            if l[4] :
                img.draw_rectangle(l[0],l[1],l[2],l[3], color=(0, 255, 0))
                img.draw_string(l[0],l[1]-24, "with mask", color=(0, 255, 0), scale=2)
            else:
                img.draw_rectangle(l[0],l[1],l[2],l[3], color=(255, 0, 0))
                img.draw_string(l[0],l[1]-24, "without mask", color=(255, 0, 0), scale=2)

    img.draw_string(0, 0, "%2.1ffps" %(fps), color=(0, 60, 128), scale=2.0)
    lcd.display(img)

kpu.deinit()

Experiment Preparation

    1. First transfer model files to the TF card, then insert the TF card into the K210 module's TF card slot. For detailed steps, see Transfer Model Files to TF Card.
    1. Prepare a face image without a mask and a face image wearing a mask.
    1. Connect the K210 to your computer via USB.
    1. Open CanMV IDE and run the routine code above.

Experiment Results

    1. After system initialization completes, the LCD displays the camera feed. Point the camera at a face. When no mask is worn, a red box and "without mask" are displayed.
    1. When a mask is worn, a green box and "with mask" are displayed.
    1. Face-related information is also printed in the Serial Terminal at the bottom of the IDE.
    1. The current mask detection threshold is threshold=0.7. Adjust this value if you need more accurate face detection.

Routine Code Explanation

    1. Import the required libraries and initialize the camera and LCD display.
import sensor, image, time, lcd from maix import KPU

lcd.init()
sensor.reset()
sensor.set_pixformat(sensor.RGB565) 
sensor.set_framesize(sensor.QVGA)
sensor.skip_frames(time = 1000) 
clock = time.clock()
    1. Initialize KPU parameters. The KPU must load a .kmodel file. This experiment uses /sd/KPU/face_mask_detect/detect_5.kmodel and YOLO2 for inference. od_img is the neural network input image at 320×256, used to store camera frames for KPU computation.
od_img = image.Image(size=(320,256), copy_to_fb=False)
anchor = (0.156250, 0.222548, 0.361328, 0.489583, 0.781250, 0.983133, 1.621094,
1.964286, 3.574219, 3.94000)
kpu = KPU()
print("ready load model")
kpu.load_kmodel("/sd/KPU/face_mask_detect/detect_5.kmodel")
kpu.init_yolo2(anchor, anchor_num=5, img_w=320, img_h=240, net_w=320 , net_h=256
,layer_w=10 ,layer_h=8, threshold=0.7, nms_value=0.4, classes=2)
    1. Create a while loop to pass images to the KPU for computation. Using the YOLO2 neural network, faces wearing masks are outlined in green with "with mask"; faces without masks are outlined in red with "without mask".
while True:
    clock.tick()
    img = sensor.snapshot()
    od_img.draw_image(img, 0,0)
    od_img.pix_to_ai()
    kpu.run_with_output(od_img)
    dect = kpu.regionlayer_yolo2()
    fps = clock.fps()
    if len(dect) > 0:
        print("dect:", dect)
        for l in dect :
            if l[4] :
                img.draw_rectangle(l[0],l[1],l[2],l[3], color=(0, 255, 0))
                img.draw_string(l[0],l[1]-24, "with mask", color=(0, 255, 0), scale=2)
            else:
                img.draw_rectangle(l[0],l[1],l[2],l[3], color=(255, 0, 0))
                img.draw_string(l[0],l[1]-24, "without mask", color=(255, 0, 0), scale=2)

    img.draw_string(0, 0, "%2.1ffps" %(fps), color=(0, 60, 128), scale=2.0)
    lcd.display(img)

kpu.deinit()
Copyright © 2026 wmnologo