Data processing and data set preparation

---

One 、 Batch renaming of pictures

• Folder structure ：
  • source
    • animals
      • cat
      • cattle
      • dog
      • horse
      • pig
    • fruits
      • apple
      • banana
      • durian
      • grape
      • orange
    • vehicles
      • bus
      • car
      • plane
      • ship
      • train

• Implementation code ：

import os
import cv2 as cv

sourcePath  = '../../../DataSet/source/'

animalPath  = sourcePath + 'animals/'
fruitPath   = sourcePath + 'fruits/'
vehiclePath = sourcePath + 'vehicles/'


#  Rename the fruit picture 
for folder_list in os.listdir(fruitPath):

    count = 1

    #  Read the subfolders under each directory 
    subfolder = fruitPath + folder_list + '/'

    for file in os.listdir(subfolder):

        old_name = subfolder + file
        new_name = subfolder + "%s_%d.jpg" % (folder_list, count)
        print(old_name, "====>", new_name)

        os.rename(old_name, new_name)
        count = count + 1


#  Animal picture rename 
for folder_list in os.listdir(animalPath):

    count = 1

    #  Read the subfolders under each directory 
    subfolder = animalPath + folder_list + '/'

    for file in os.listdir(subfolder):

        old_name = subfolder + file
        new_name = subfolder + "%s_%d.jpg" % (folder_list, count)

        print(old_name, "====>", new_name)

        os.rename(old_name, new_name)
        count = count + 1


#  Traffic picture rename 
for folder_list in os.listdir(vehiclePath):

    count = 1

    #  Read the subfolders under each directory 
    subfolder = vehiclePath + folder_list + '/'

    for file in os.listdir(subfolder):

        old_name = subfolder + file
        new_name = subfolder + "%s_%d.jpg" % (folder_list, count)

        print(old_name, "====>", new_name)

        os.rename(old_name, new_name)
        count = count + 1

• Realization effect ：
  

---

Two 、 Data to enhance + Gaussian noise

During the game, the camera is used to capture images for recognition , The captured image contains a lot of noise , The original data set is the original image without noise . For example, a picture of a dog is affected by indoor light , There's a lot of red noise , It is easy to be recognized as apple . This happened last year , Therefore, this year is aimed at solving , Add noise to the data set .

• Implementation code ：

#  Image denoising 
#  There will be a lot of noise in the actual pictures 
#  There is a big gap with the original computer drawing 
#  Therefore, it is necessary to add noise to the original image 

import numpy as np
import cv2   as cv
import os
from keras.preprocessing.image import ImageDataGenerator
from keras.preprocessing.image import img_to_array
from keras.preprocessing.image import load_img


#  Noise adding function 
def Gasuss_Noise(image, mean=0, var=0.001):
    '''  Add Gaussian noise  image:  original image  mean :  mean value  var :  variance ,  The bigger it is , The louder the noise  '''
    image = cv.imread(image)
    image = np.array(image/255, dtype=float)                                                                            #  Normalize the pixel value of the original image , Divide 255 Make the pixel value in 0-1 Between 
    noise = np.random.normal(mean, var ** 0.5, image.shape)                                                             #  Create an average of mean, The variance of var Image matrix with Gaussian distribution 
    out = image + noise                                                                                                 #  Add the noise and the original image to get the noisy image 
    if out.min() < 0:
        low_clip = -1.
    else:
        low_clip = 0.
    out = np.clip(out, low_clip, 1.0)                                                                                   # clip The function limits the size of the element to low_clip and 1 Between , Less than low_clip Instead of , Greater than 1 The use of 1 Instead of 
    out = np.uint8(out * 255)                                                                                             #  De normalize , multiply 255 Restore the pixel value of the noisy image 
    # cv.imshow("gasuss", out)
    noise = noise * 255
    return [noise, out]


#  Image enhancement function 
def Date_Enhancement(img_input_path, img_output_path):
    image = load_img(img_input_path)

    im1 = image.point(lambda p: p * 0.6)
    # im1.show()

    im1 = img_to_array(im1)  #  Image to array 
    im1 = np.expand_dims(im1, axis=0)  #  Add a dimension 

    img_dag = ImageDataGenerator(
        rotation_range=10,
        width_shift_range=0.001,   # horizen offset
        height_shift_range=0.001,  # vertical offset
        shear_range=0.02,
        # zoom_range=[0.6, 0.9],
        brightness_range = [0.9, 1.1],
        horizontal_flip=False,    #  Flip horizontal 
        fill_mode="constant", cval=40
    )  #  rotate , Width movement range , Height movement range , Clipping range , Flip horizontally to open , Fill mode 

    img_generator = img_dag.flow(im1,
                                 batch_size=1,
                                 save_to_dir= img_output_path,
                                 save_prefix= "image",
                                 save_format= "jpg")  #  Test an image bath_size=1
    count = 0  #  Counter 

    for raw_pic_path in img_generator:
        count += 1
        if count == 1:  #  How many samples are generated before exiting 
            break


####################################################################

#  The original address 
sourcePath  = '../../../DataSet/source/'
animalPath  = sourcePath + 'animals/'
fruitPath   = sourcePath + 'fruits/'
vehiclePath = sourcePath + 'vehicles/'

#  Intermediate address 
#  It is used to save the picture after the first step 
middlePath        = '../../../DataSet/middle/'
animalMiddlePath  = middlePath + 'animals/'
fruitMiddlePath   = middlePath + 'fruits/'
vehicleMiddlePath = middlePath + 'vehicles/'

#  Destination address 
#  Save the image data processed in the second step 
processPath        = '../../../DataSet/process/'
animalProcessPath  = processPath + 'animals/'
fruitProcessPath   = processPath + 'fruits/'
vehicleProcessPath = processPath + 'vehicles/'


for folder_name in os.listdir(fruitPath):

    rename_count = 1

    #  Read the subfolders under each directory 
    sub_folder = fruitPath + folder_name + '/'
    middle_save_path = fruitMiddlePath + folder_name + '/'

    #  Data to enhance 
    for raw_pic in os.listdir(sub_folder):
        raw_pic_path = sub_folder + raw_pic
        for i in range(0, 10):
            Date_Enhancement(raw_pic_path, middle_save_path)

    print(folder_name, "Enhance Done")

    #  Gaussian noise 
    for middle_pic in os.listdir(middle_save_path):
        middle_save_name = middle_save_path + middle_pic
        final_save_path = fruitProcessPath + folder_name + '/'
        final_save_name = final_save_path + "%s_%d.jpg" % (folder_name, rename_count)

        noise, out = Gasuss_Noise(middle_save_name, mean=0, var=0.002)

        cv.imwrite(final_save_name, out)
        rename_count = rename_count + 1

    print(folder_name, "Process Done")


for folder_name in os.listdir(animalPath):

    rename_count = 1

    #  Read the subfolders under each directory 
    sub_folder = animalPath + folder_name + '/'
    middle_save_path = animalMiddlePath + folder_name + '/'

    #  Data to enhance 
    for raw_pic in os.listdir(sub_folder):
        raw_pic_path = sub_folder + raw_pic
        for i in range(0, 10):
            Date_Enhancement(raw_pic_path, middle_save_path)

    print(folder_name, "Enhance Done")

    #  Gaussian noise 
    for middle_pic in os.listdir(middle_save_path):
        middle_save_name = middle_save_path + middle_pic
        final_save_path = animalProcessPath + folder_name + '/'
        final_save_name = final_save_path + "%s_%d.jpg" % (folder_name, rename_count)

        noise, out = Gasuss_Noise(middle_save_name, mean=0, var=0.002)

        cv.imwrite(final_save_name, out)
        rename_count = rename_count + 1

    print(folder_name, "Process Done")


for folder_name in os.listdir(vehiclePath):

    rename_count = 1

    #  Read the subfolders under each directory 
    sub_folder = vehiclePath + folder_name + '/'
    middle_save_path = vehicleMiddlePath + folder_name + '/'

    #  Data to enhance 
    for raw_pic in os.listdir(sub_folder):
        rename_count = 1
        raw_pic_path = sub_folder + raw_pic
        for i in range(0, 10):
            Date_Enhancement(raw_pic_path, middle_save_path)

    print(folder_name, "Enhance Done")

    #  Gaussian noise 
    for middle_pic in os.listdir(middle_save_path):
        middle_save_name = middle_save_path + middle_pic
        final_save_path = vehicleProcessPath + folder_name + '/'
        final_save_name = final_save_path + "%s_%d.jpg" % (folder_name, rename_count)

        noise, out = Gasuss_Noise(middle_save_name, mean=0, var=0.002)

        cv.imwrite(final_save_name, out)
        rename_count = rename_count + 1

    print(folder_name, "Process Done")

---

3、 ... and 、 Data set preparation

【 explain 】python access files / Folders are sorted alphabetically , So the order of labels is ：

• Implementation code ：

import os
import cv2 as cv
import numpy as np

make = True
check = True

#  Picture address 
picPath        = '../../../DataSet/process/'

if __name__ == "__main__":

    if make:
        all_data = []
        all_label = []

        i = -1
        for fruit_animal_vehicle in os.listdir(picPath):
            for apple_cat_bus in os.listdir(picPath + fruit_animal_vehicle + '/'):
                i = i + 1
                for pic in os.listdir(picPath + fruit_animal_vehicle + '/' + apple_cat_bus + '/'):

                    extension = os.path.splitext(pic)[-1]
                    if extension == '.jpg':
                        img = cv.imread(picPath + fruit_animal_vehicle + '/' + apple_cat_bus + '/' + pic)  #  Read image data to img ## Connect two pathname components  # cv2.imread  The image data read is  BGR  Format ;
                        try:                                                   #  If try Something goes wrong 
                            img = cv.resize(img, (32, 32))[..., (2, 1, 0)]        # BGR 2 RGB
                            all_data.append(img)                              # .append Method adds... To the end of the array img
                            all_label.append(i)                               # .append Method 
                        except:                                                #  perform except
                            continue

        all_data  = np.asarray(all_data)
        all_label = np.asarray(all_label)

        np.save("../../../DataSet/dataSet/pic", all_data)
        np.save("../../../DataSet/dataSet/label", all_label)

    if check:
        x = np.load("../../../DataSet/dataSet/pic.npy")
        y = np.load("../../../DataSet/dataSet/label.npy")

        label = ["cat", "cattle", "dog", "horse", "pig"] \
              + ["apple", "banana", "durian", "grape", "orange"] \
              + ["bus", "car", "plane", "ship", "train"]

        count = 0
        for d, idx in zip(x, y):                                             ##  Package the corresponding elements in the object into tuples , Then return a list of these tuples 
            print("Class %s %d" % (label[idx], count))
            d = cv.resize(d, (395, 395))[..., (2, 1, 0)]
            cv.imshow("img", d)
            count = count + 1
            cv.waitKey(0)