Merge pull request #220 from tukcomCD2024/AI#36/Preprocessing

Ai#36/preprocessing

.gitignore

@@ -1,5 +1,7 @@
 
+
+*.h5
 .env
 
 backend/shareNote/src/main/resources/application-email.yml
-*.h5
+

AI/Model/ImageDataArrange.py

@@ -0,0 +1,30 @@
+import os
+import shutil
+
+
+def getImageListByHash(path):
+    img = [{}]  # imagesWithTag
+    for i in os.listdir(path):
+        if os.path.isdir(f'{path}/{i}'):
+            img[0][i] = getImageListByHash(f'{path}/{i}')
+        else:
+            img.append(i)
+    if len(img[0]) == 0:
+        img.pop(0)
+    if len(img) == 1:
+        img = img[0]
+    return img
+
+
+# print(getImageListByHash('../asset/image/sample/svg/'))
+
+def imageCategorical():
+    src = 'C:/Users/Ka/Desktop/Ka/programming/AI/AI2/asset/size224Image01'
+    for routes, dirs, files in os.walk(src):
+        print(routes, files, dirs)
+        for image in files:
+            os.mkdir(routes + '/' + image[:-4])
+            shutil.move(routes + '/' + image, routes + '/' + image[:-4] + '/' + image)
+
+
+imageCategorical()

AI/Model/ImageDataGenerator.py

@@ -0,0 +1,54 @@
+from keras.preprocessing.image import ImageDataGenerator
+import matplotlib.pylab as plt
+
+data_file_path = r"C:\Users\Ka\Desktop\Ka\대학교\졸업작품\project\Code-Bridge\AI\asset\image"
+
+train_data_generator = ImageDataGenerator(
+    rescale=1. / 255,
+    featurewise_center=False,  # set input mean to 0 over the dataset
+    samplewise_center=False,  # set each sample mean to 0
+
+    featurewise_std_normalization=False,  # divide inputs by std of the dataset
+    samplewise_std_normalization=False,  # divide each input by its std
+    zca_whitening=False,  # apply ZCA whitening
+    rotation_range=0,  # randomly rotate images in the range (degrees, 0 to 180)
+    width_shift_range=0.2,  # randomly shift images horizontally (fraction of total width)
+    height_shift_range=0.2,  # randomly shift images vertically (fraction of total height)
+    zoom_range=[0.7, 1.1],
+    horizontal_flip=True,  # randomly flip images
+    vertical_flip=False,
+    validation_split=0.1)  # randomly flip images
+
+test_data_generator = ImageDataGenerator(rescale=1. / 255)
+
+train_generator = train_data_generator.flow_from_directory(
+    data_file_path + "/animals",
+    target_size=(224, 224),
+    batch_size=3,
+    class_mode='categorical',
+    subset='training')
+
+test_generator = train_data_generator.flow_from_directory(
+    data_file_path + "/animals",
+    target_size=(224, 224),
+    batch_size=3,
+    class_mode='categorical',
+    subset='validation')
+
+# x_train, y_train = train_generator.next()
+# for idx in range(len(x_train)):
+#     print(x_train[idx].shape)
+#     print(y_train[idx])
+#     plt.imshow(x_train[idx])
+#     plt.show()
+#
+# print(len(x_train))
+# print(train_generator.n)
+#
+# for i in range(40):
+#     x_train, y_train = train_generator.next()
+#     for idx in range(len(x_train)):
+#         # print(x_train[idx].shape)
+#         # print(y_train[idx])
+#         plt.imshow(x_train[idx])
+#         plt.show()

AI/Model/cnnModel.py

@@ -0,0 +1,99 @@
+from keras.models import Sequential
+from keras.layers import Dense, Conv2D, MaxPool2D, Flatten
+from keras.preprocessing.image import ImageDataGenerator
+from keras.callbacks import ModelCheckpoint, EarlyStopping
+from keras.optimizers import RMSprop
+import os
+
+os.environ['KMP_DUPLICATE_LIB_OK'] = 'True'
+imageSize = 128
+
+trdata = ImageDataGenerator(zoom_range=[0.9, 1.3], height_shift_range=0.3, width_shift_range=0.2, horizontal_flip=True, rotation_range=0.3,
+                            validation_split=0.1)
+traindata = trdata.flow_from_directory(directory=r"C:\Users\Ka\Desktop\Ka\대학교\졸업작품\project\Code-Bridge\AI\asset\image\svg",
+                                       target_size=(imageSize, imageSize), class_mode='categorical', batch_size=20)
+tsdata = ImageDataGenerator()
+testdata = tsdata.flow_from_directory(directory=r"C:\Users\Ka\Desktop\Ka\대학교\졸업작품\project\Code-Bridge\AI\asset\image\svg", target_size=(imageSize, imageSize),
+                                      class_mode='categorical')
+
+
+def cnnDepth4():
+    model = Sequential()
+    model.add(Conv2D(32, (3, 3), activation='relu', input_shape=(imageSize, imageSize, 3)))
+    model.add(MaxPool2D((2, 2)))
+    model.add(Conv2D(64, (3, 3), activation='relu'))
+    model.add(MaxPool2D((2, 2)))
+
+    model.add(Flatten())
+    model.add(Dense(units=1024, activation="relu"))
+    model.add(Dense(units=103, activation="sigmoid"))
+
+    return model
+
+
+def cnnDepth5():
+    model = Sequential()
+    model.add(Conv2D(32, (3, 3), activation='relu', input_shape=(imageSize, imageSize, 3)))
+    model.add(MaxPool2D((2, 2)))
+    model.add(Conv2D(64, (3, 3), activation='relu'))
+    model.add(MaxPool2D((2, 2)))
+
+    model.add(Flatten())
+    model.add(Dense(units=1024, activation="relu"))
+    model.add(Dense(units=1024, activation="relu"))
+    model.add(Dense(units=103, activation="sigmoid"))
+
+    return model
+
+
+def cnnDepth6():
+    model = Sequential()
+    model.add(Conv2D(32, (3, 3), activation='relu', input_shape=(imageSize, imageSize, 3)))
+    model.add(MaxPool2D((2, 2)))
+    model.add(Conv2D(64, (3, 3), activation='relu'))
+    model.add(MaxPool2D((2, 2)))
+    model.add(Conv2D(64, (3, 3), activation='relu'))
+    model.add(MaxPool2D((2, 2)))
+
+    model.add(Flatten())
+    model.add(Dense(units=1024, activation="relu"))
+    model.add(Dense(units=1024, activation="relu"))
+    model.add(Dense(units=103, activation="sigmoid"))
+
+    return model
+
+
+def createModel(model, lr, e1, name):
+    modelNames = ['cnn4Depth.h5', 'cnn5Depth.h5', 'cnn6Depth.h5']
+    modelFuncs = [cnnDepth4(), cnnDepth5(), cnnDepth6()]
+
+    opt = RMSprop(lr=lr)
+    model.compile(optimizer=opt, loss='categorical_crossentropy', metrics=['accuracy'])
+
+    model.summary()
+
+    checkpoint = ModelCheckpoint(name, monitor='val_accuracy', verbose=1, save_best_only=True,
+                                 save_weights_only=False, mode='auto', period=1)
+    early = EarlyStopping(monitor='val_accuracy', min_delta=0, patience=30, verbose=1, mode='auto')
+    # hist = model.fit_generator(steps_per_epoch=len(traindata), generator=traindata, validation_data=testdata,
+    #                            validation_steps=len(testdata), epochs=4, callbacks=[checkpoint, early])
+
+    hist = model.fit(traindata, steps_per_epoch=len(traindata), validation_data=testdata, validation_steps=len(testdata),
+                     epochs=e1, callbacks=[checkpoint, early], batch_size=5)
+
+    import matplotlib.pyplot as plt
+
+    plt.plot(hist.history["accuracy"])
+    plt.plot(hist.history['val_accuracy'])
+    plt.plot(hist.history['val_loss'])
+    plt.title("model accuracy")
+    plt.ylabel("Accuracy")
+    plt.xlabel("Epoch")
+    plt.ylim(0, 5)
+    plt.legend(["Accuracy", "Validation Accuracy", "Validation Loss"])
+    plt.show()
+
+for e in range(40, 110, 10):
+    createModel(cnnDepth6(), 0.0001, e, f'cnn6f32e{e}u1024.h5')
+    createModel(cnnDepth5(), 0.0001, e, f'cnn5f32e{e}u1024.h5')
+    createModel(cnnDepth4(), 0.0001, e, f'cnn4f32e{e}u1024.h5')

AI/Model/test.py

@@ -0,0 +1,47 @@
+import os
+from PIL import Image
+
+import numpy as np
+
+
+def resultByDesc(result):
+    x = {}
+    for i in range(len(result)):
+        if result[i] > 0.001:
+            x[i] = result[i]
+
+    import operator
+    return sorted(x.items(), key=operator.itemgetter(1), reverse=True)[:40]
+
+
+# imgsrc = r"C:\Users\Ka\Desktop\Ka\programming\AI\AI2\asset\size64Image01\bank\bank.png"
+# imgsrc = r"C:\Users\Ka\Desktop\Ka\programming\AI\AI2\asset\size64Image01\security\security.png"
+imgsrc = r"C:\Users\Ka\Desktop\Ka\programming\AI\sample\{}.png"
+images = {'arrow': 4, 'heart': 51, 'heart2': 51, 'cloud': 28, 'sword': 91, 'skirt': 83, 'star': 87, 'star2': 87, 'bread':17, 'bread2':17, 'candy':26}
+
+from keras.models import load_model
+
+for i in os.listdir('./'):
+    if not 'f32' in i:
+        continue
+    print(i)
+
+    saved_model = load_model(i)
+
+    for i in images.keys():
+        img = Image.open(imgsrc.format(i))
+        img = img.resize((128, 128))
+        img = img.convert("RGB")
+        img = np.array(img)
+        img = np.expand_dims(img, axis=0)
+
+        output = saved_model.predict(img)
+
+        result = resultByDesc(output[0])[:8]
+        for j in range(5):
+            try:
+                if images[i] in result[j]:
+                    print(i, j)
+                    break
+            except:
+                break

AI/Preprocessing/CheckImageExist.py

@@ -0,0 +1,46 @@
+import csv
+import chromedriver_autoinstaller as cd_installer
+from selenium import webdriver
+from selenium.webdriver.common.by import By
+
+cd_installer.install()
+
+
+def getReaderTXT(url, token):
+    f = open(url)
+    return f.readline().split(token)
+
+
+def getReaderCSV(url):
+    f = open(url)
+    return csv.reader(f)
+
+def availableTag(reader):
+    driver = webdriver.Chrome()
+    checkMsgPath = '//*[@id="app"]/div[2]/div/div/div[2]/div[1]'
+
+    f = open('../asset/tag/tagsFinal5.txt', 'w')
+    print(len(reader))
+    for row in reader:
+        driver.get(f"https://icon-sets.iconify.design/?query={row}")
+        driver.implicitly_wait(200)
+        try:
+            el = driver.find_element(By.XPATH, checkMsgPath)
+            if el.text != 'No icon sets match your search':
+                f.write(row+"\n")
+        except Exception as e:
+            pass
+
+    f = open('../asset/tag/sampleTag.txt', 'r')
+
+    availableTags = []
+    for i in f.readlines():
+        availableTags.append(i[:-1])
+    writer = csv.DictWriter(open('../asset/tag/sampleTags.csv', 'w', newline='\n'), fieldnames=['tag', 'count'])
+    writer.writeheader()
+    for row in reader:
+        if row[0] in availableTags:
+            writer.writerow({'tag': row[0], 'count': row[1]})
+
+
+availableTag(getReaderTXT('../asset/tag/sampleTag.txt', ', '))