Changing from tf module to ml module

Project Discussion
1

Hi, I want to write a code for face with emotion detection. I am unsure how to modify the code as I was using the tf module and I understand that the module is now gone. Below is my current code. Please advice.

import sensor, image, time, tf

sensor.reset()
sensor.set_pixformat(sensor.RGB565)
sensor.set_framesize(sensor.QVGA)
sensor.skip_frames(time = 2000)

clock = time.clock()

net = tf.Model("trained.tflite", load_to_fb = True)
labels = [l.rstrip('\n') for l in open ("labels.txt")]

while(True):
    clock.tick()

    #Take a picture and brighten things up for the frontal face detector.
    img = sensor.snapshot().gamma_corr(contrast = 1.5)

    #Returns a list of rects (x,y,w,h) where faces are.
    faces = img.find_features(image.HaarCascade("frontalface"))

    for f in faces:

        #Classify a face and get the class scores list
        scores = net.predict(img, roi = f)[0].output()

        #Find the highest class score and lookup the label for that
        label = labels[scores.index(max(scores))]

        #Draw a box around the face
        img.draw_rectangle(f)

        #Draw the label above the face
        img.draw_string(f[0]+3, f[1]-1, label, mono_space = False)

    print(clock.fps())
2

Hi, I wrote a massive post about this:

OpenMV Firmware v4.5.6 and up TensorFlow Porting Guide - OpenMV Products - OpenMV Forums

3

Do:

import sensor, image, time, ml

sensor.reset()
sensor.set_pixformat(sensor.RGB565)
sensor.set_framesize(sensor.QVGA)
sensor.skip_frames(time = 2000)

clock = time.clock()

net = ml.Model("trained.tflite", load_to_fb = True)
labels = [l.rstrip('\n') for l in open ("labels.txt")]

while(True):
    clock.tick()

    #Take a picture and brighten things up for the frontal face detector.
    img = sensor.snapshot().gamma_corr(contrast = 1.5)

    #Returns a list of rects (x,y,w,h) where faces are.
    faces = img.find_features(image.HaarCascade("frontalface"))

    for f in faces:

        norm = ml.Normalization(roi=f)
        #Classify a face and get the class scores list
        scores = net.predict([norm(img)])[0].flatten().tolist()

        #Find the highest class score and lookup the label for that
        label = labels[scores.index(max(scores))]

        #Draw a box around the face
        img.draw_rectangle(f)

        #Draw the label above the face
        img.draw_string(f[0]+3, f[1]-1, label, mono_space = False)

    print(clock.fps())

While the normalization object is a bit awkward, it’s the only way to allow for multi-input networks to have arguments per input.

In the future you will be able to have a net operating on camera and distance sensor info at the same time and etc.

You could also do image crops too:

import sensor, image, time, ml

sensor.reset()
sensor.set_pixformat(sensor.RGB565)
sensor.set_framesize(sensor.QVGA)
sensor.skip_frames(time = 2000)

clock = time.clock()

net = ml.Model("trained.tflite", load_to_fb = True)
labels = [l.rstrip('\n') for l in open ("labels.txt")]

while(True):
    clock.tick()

    #Take a picture and brighten things up for the frontal face detector.
    img = sensor.snapshot().gamma_corr(contrast = 1.5)

    #Returns a list of rects (x,y,w,h) where faces are.
    faces = img.find_features(image.HaarCascade("frontalface"))

    for f in faces:

        #Classify a face and get the class scores list
        scores = net.predict([img.crop(roi=f, copy=True)])[0].flatten().tolist()

        #Find the highest class score and lookup the label for that
        label = labels[scores.index(max(scores))]

        #Draw a box around the face
        img.draw_rectangle(f)

        #Draw the label above the face
        img.draw_string(f[0]+3, f[1]-1, label, mono_space = False)

    print(clock.fps())

However, this is slower and wastes a lot of RAM.

1 Like