Training ConvNext Model

Capstone_DVD/Train_7_category_lists/ConvNextTesting.py

@@ -0,0 +1,189 @@
+import keras
+from keras.applications.convnext import ConvNeXtBase
+import numpy as np
+import tensorflow as tf
+from sklearn.preprocessing import LabelBinarizer
+import cv2
+from csv import reader
+import os
+import argparse
+from random import shuffle
+
+
+def step_decay(epoch):
+    initAlpha=.0020
+    if epoch <= 20:
+        alpha=initAlpha
+    elif epoch <= 40:
+        alpha=initAlpha*0.75
+    elif epoch <= 60:
+        alpha=initAlpha*0.50
+    elif epoch <= 80:
+        alpha=initAlpha*0.25
+    elif epoch <=100:
+        alpha=initAlpha*0.175
+    elif epoch <= 125:
+        alpha=initAlpha*0.10
+    else:
+        alpha=initAlpha*0.05
+    return float(alpha)
+
+
+def create_data_arrays_list(imageData, image_dimension, verbose=-1):
+    # initialize the list of features and labels
+    data = []
+    labels = []
+    i=0
+
+    # loop over the input images
+    for image_input in imageData:
+        # load the image
+        images_dir = os.getcwd() + "/image_data/images/"
+        image = cv2.imread(images_dir + image_input[0][:])
+
+
+
+        #image = cv2.convertScaleAbs(image, alpha=alpha, beta=beta)
+
+
+        # resize image, convert image to grayscale, then back to original color scheme
+        image = cv2.cvtColor(cv2.cvtColor(cv2.resize(image, (image_dimension, image_dimension),
+            interpolation=cv2.INTER_AREA),cv2.COLOR_BGR2GRAY),cv2.COLOR_GRAY2BGR)
+
+        #convert image into array format using Keras function
+        image=keras.utils.img_to_array(image)
+
+        #add image data to array
+        data.append(image)
+
+        #Change label names so that class numbers will correspond to chronological sequence; class names changed back further on
+        if image_input[:][1] == 'Kanaa': append_label='00'+'Kanaa'
+        elif image_input[:][1] == 'Black_Mesa': append_label='01'+'Black_Mesa'
+        elif image_input[:][1] == 'Sosi': append_label='02'+'Sosi'
+        elif image_input[:][1] == 'Dogoszhi': append_label='03'+'Dogoszhi'
+        elif image_input[:][1] == 'Flagstaff': append_label='04'+'Flagstaff'
+        elif image_input[:][1] == 'Tusayan': append_label='05'+'Tusayan'
+        else: append_label='06'+'Kayenta'
+
+        #write label name
+        labels.append(append_label)
+
+        # show an update every `verbose` images
+        if verbose > 0 and i > 0 and (i + 1) % verbose == 0:
+            print("[INFO] processed {}/{}".format(i + 1,
+            len(imageData)))
+        i=i+1
+
+    # return image array data, labels
+    return (np.array(data), np.array(labels))
+
+# Arg parser for set number
+ap = argparse.ArgumentParser()
+ap.add_argument("-s", "--set", required=True,
+	help="integer index of sherd train/test set")
+args = vars(ap.parse_args())
+
+set_number=str(args["set"])
+set_directory= os.getcwd() + "/image_data/"
+images_dir = os.getcwd() + "/image_data/"
+num_classes = 7
+image_dimension = 224
+batch_size = 8
+epochs_head = 10
+epochs = 90
+l2_constant=0.02
+input = keras.Input(shape=(image_dimension, image_dimension, 3))
+
+#Define directories to use based on set number for loading data, saving data
+train_dataset=set_directory + "/Set_" + set_number +"/train_" + set_number + ".csv"
+test_dataset=set_directory + "/Set_" + set_number +"/test_" + set_number + ".csv"
+
+# Get lists of train images, types from file
+print("[INFO] loading images...")
+with open(train_dataset, 'r') as read_obj:
+    # pass the file object to reader() to get the reader object
+    csv_reader = reader(read_obj)
+    # Pass reader object to list() to get a list of lists
+    train_data_list = list(csv_reader)
+
+#close file
+read_obj.close()
+
+#randomize order of images
+shuffle(train_data_list)
+
+#same for test images, types
+with open(test_dataset, 'r') as read_obj:
+    # pass the file object to reader() to get the reader object
+    csv_reader = reader(read_obj)
+    # Pass reader object to list() to get a list of lists
+    test_data_list = list(csv_reader)
+
+
+read_obj.close()
+
+(train_data,train_labels)=create_data_arrays_list(train_data_list,image_dimension, verbose=250)
+
+(test_data,test_labels)=create_data_arrays_list(test_data_list,image_dimension,verbose=250)
+
+#Set classNames from train_labels list, rename to eliminate numbers from the front
+classNames = [str(x) for x in np.unique(train_labels)]
+classNames = ['Kanaa',  'Black_Mesa', 'Sosi', 'Dogoszhi', 'Flagstaff', 'Tusayan', 'Kayenta']
+
+#(x_train, y_train), (x_test, y_test) = keras.datasets.cifar10.load_data()
+
+(x_train, x_test, y_train, y_test) = (train_data,test_data,train_labels,test_labels)
+
+# convert the labels from integers to vectors
+y_train = LabelBinarizer().fit_transform(y_train)
+y_test = LabelBinarizer().fit_transform(y_test)
+
+
+baseModel = ConvNeXtBase(include_top=False, input_tensor=input)
+
+headModel = baseModel.output
+headModel = keras.layers.GlobalAveragePooling2D()(headModel)
+headModel = keras.layers.Flatten()(headModel)
+headModel = keras.layers.Dense(1024, activation="relu", 
+                               kernel_regularizer=keras.regularizers.L2(l2_constant))(headModel)
+headModel = keras.layers.Dense(num_classes, activation="softmax")(headModel)
+
+model = keras.models.Model(inputs=baseModel.input, outputs=headModel)
+
+
+for layer in baseModel.layers:
+   layer.trainable = False
+
+opt = keras.optimizers.RMSprop(learning_rate=0.01)
+
+model.compile(loss="categorical_crossentropy", optimizer=opt,
+              metrics=["accuracy"])
+
+#Add Learning rate scheduler to model callbacks
+callbacks=[keras.callbacks.LearningRateScheduler(step_decay)]
+
+print("[INFO] training head...")
+
+model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs_head, 
+          validation_split=0.1, validation_data=(x_test, y_test), 
+          callbacks=callbacks)
+
+for layer in baseModel.layers[0:]:
+   layer.trainable = True
+
+print("Re-compiling model...")
+
+# Have to rebuild optimizer for model recompile
+opt = keras.optimizers.SGD(learning_rate=0.005)
+model.compile(loss="categorical_crossentropy", optimizer=opt,
+              metrics=["accuracy"])
+print("Model Compiled")
+
+print("Fine-Tuning Final Model...")
+model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs,
+          validation_split=0.1, validation_data=(x_test, y_test), 
+          callbacks=callbacks)
+
+score = model.evaluate(x_test, y_test, verbose=1)
+print("Test loss:", score[0])
+print("Test accuracy:", score[1])

Capstone_DVD/Train_7_category_lists/ResNet152V2_model_processor_lists.py

@@ -6,9 +6,9 @@
 from keras.applications import ResNet152, ResNet152V2,DenseNet169, ResNet50,resnet,VGG16,vgg16,resnet_v2
 from keras.layers import Input, Dropout,Flatten,Dense,GlobalAveragePooling2D
 from keras.models import Model,load_model
-from keras.preprocessing.image import ImageDataGenerator
-from tensorflow.keras.utils import img_to_array
-from tensorflow.keras.applications import imagenet_utils
+from keras.src.legacy.preprocessing.image import ImageDataGenerator
+from keras.utils import img_to_array
+from keras.applications import imagenet_utils
 from imutils import paths
 from random import shuffle
 import numpy as np
@@ -21,7 +21,7 @@
 
 
 #Needed for top layers of model
-from tensorflow.keras.layers import BatchNormalization
+from keras.layers import BatchNormalization
 from keras import regularizers
 
 #Adds top classification layer to transfer learning model