텐서플로우 - 간단한 기계학습 시키기

 

https://teachablemachine.withgoogle.com/

Teachable Machine

 

여러가지 종류의 프로젝트를 생성할 수 있음

 

 

[이미지 프로젝트 로컬에서 실행시키기]

import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'

import tensorflow as tf
from keras.models import load_model
from PIL import Image, ImageOps
import numpy as np

import cv2

capture = cv2.VideoCapture(0)
capture.set(cv2.CAP_PROP_FRAME_WIDTH, 640)
capture.set(cv2.CAP_PROP_FRAME_HEIGHT, 480)

label_class = { 0: '포스터잇',
                   1: '스마트폰',
                   2: '리모콘' }

# Load the model
model = load_model('./keras_model.h5')

# Create the array of the right shape to feed into the keras model
# The 'length' or number of images you can put into the array is
# determined by the first position in the shape tuple, in this case 1.
data = np.ndarray(shape=(1, 224, 224, 3), dtype=np.float32)
# Replace this with the path to your image

while cv2.waitKey(33) < 0:
    try:
        ret, frame = capture.read()
        cv2.imshow("VideoFrame", frame)

        size = (224, 224)

        frame = cv2.resize(frame, size)
        image = frame

        #turn the image into a numpy array
        image_array = np.asarray(image)
        # Normalize the image
        normalized_image_array = (image_array.astype(np.float32) / 127.0) - 1
        # Load the image into the array
        data[0] = normalized_image_array

        # run the inference
        predictions = model.predict(data)
        score = tf.nn.softmax(predictions[0])

        print( label_class[np.argmax(score)], 100 * np.max(score) )
    except:
        None



        



capture.release()
cv2.destroyAllWindows()

 

[좀더 정확한 버전]

import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'

import tensorflow as tf
from keras.models import load_model
import numpy as np

import cv2
import math


label_class = { 0: 'STAY',
                   1: 'STOP',
                   2: 'UP',
				   3: 'DOWN'}

gages = {}

def initGages(label_class):
    for label in label_class.values():
        gages[ label ] = 0




max_gage = 20

def modeSelector(mode, frame):
    global gages

    for key1 in gages:

        if mode == key1:
            gages[key1] += 1
            
            for key2 in gages:
                if key2 != mode:
                    gages[key2] = 0
            
            cv2.putText(frame, key1 + ': ' + str( math.floor ( (gages[key1] / max_gage) * 100) ) + "%", (10, 50), cv2.FONT_HERSHEY_SIMPLEX, 1.5, (255, 255, 255),  3, cv2.LINE_AA)

    for key in gages:

        if gages[key] >= max_gage:
            gages[key] = 0
            cv2.putText(frame, '[' + str(key) + ']', (200, 200), cv2.FONT_HERSHEY_SIMPLEX, 2, (0, 0, 255),  3, cv2.LINE_AA)


    return frame
    
if __name__ == '__main__':
    initGages( label_class )



    capture = cv2.VideoCapture(0)
    capture.set(cv2.CAP_PROP_FRAME_WIDTH, 640)
    capture.set(cv2.CAP_PROP_FRAME_HEIGHT, 480)



    # Load the model
    model = load_model('./keras_model.h5')

    data = np.ndarray(shape=(1, 224, 224, 3), dtype=np.float32)

    while cv2.waitKey(33) < 0:
        try:
            ret, frame = capture.read()

            size = (224, 224)
            image = cv2.resize(frame, size)
 
            #turn the image into a numpy array
            image_array = np.asarray(image)
            # Normalize the image
            normalized_image_array = (image_array.astype(np.float32) / 127.0) - 1
            # Load the image into the array
            data[0] = normalized_image_array

            # run the inference
            predictions = model.predict(data)
            score = tf.nn.softmax(predictions[0])

            print( label_class[np.argmax(score)] )
            frame = modeSelector( label_class[np.argmax(score)], frame )
            
            cv2.imshow("VideoFrame", frame)

        except Exception as e:
            None

    capture.release()
    cv2.destroyAllWindows()