Routine explanation 25-Machine-Learning->nn_cifar10 neural network routine

(Note: This routine is deleted in firmware 3.6.5 and later. Use TensorFlow Lite instead, which has better results. See video tutorial 42 - Neural Network Target Detection: https://singtown.com/learn/50918/)

Video Tutorial 22 - cifar10 Neural Network: https://singtown.com/learn/50045/

Before running this example, please go to OpenMV IDE->Tools->Machine Vision->CNN Network Library and save the corresponding neural network file to the SD memory card of OpenMV.

# cifar10例程
import sensor, image, time, os, nn

sensor.reset()                         # 复位并初始化传感器。

sensor.set_contrast(3)
sensor.set_pixformat(sensor.RGB565)    # 设置像素格式为RGB565

sensor.set_framesize(sensor.QVGA)      # 将图像大小设置为QVGA (320x240)

sensor.set_windowing((128, 128))       # 设置128 x128窗口。
sensor.skip_frames(time=1000)
sensor.set_auto_gain(False)
sensor.set_auto_exposure(False)

# 加载cifar10网络。OpenMV3 M7上使用此网络可能会超出内存。
#net = nn.load('/cifar10.network')

# 更快，更小，更准确。建议OpenMV3 M7上使用此网络。
net = nn.load('/cifar10_fast.network')
labels = ['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']

clock = time.clock()                # 创建一个时钟对象来跟踪FPS帧率。
while(True):
    clock.tick()                    # 更新FPS帧率时钟。
    img = sensor.snapshot()         # 拍一张照片并返回图像。
    out = net.forward(img)
    max_idx = out.index(max(out))
    score = int(out[max_idx]*100)
    if (score < 70):
        score_str = "??:??%"
    else:
        score_str = "%s:%d%% "%(labels[max_idx], score)
    img.draw_string(0, 0, score_str, color=(255, 0, 0))

    print(clock.fps())             
    # 注意: 当连接电脑后，OpenMV会变成一半的速度。当不连接电脑，帧率会增加。
    #打印当前的帧率。

Singtown Technology OpenMV official Chinese document function explanation: