Tensorflow: Best way of creating training data with float as label - image

Question

I want to use tensorflow version 2.4.0-dev20201009 in python 3.7.
My dataset are in the subfolder "data\Images". The label of an image is a float number between 1 and 5 and can read from the allTestData.csv from the subfolder "data".
What is the best way to read the data with validation split of 30 percent? So far I wanted to use
tf.keras.preprocessing.image_dataset_from_directory but this doesn't help me to incorperate the labels correctly, as all my images are in one folder and do not have one-hot encoded vectors as labels. How would you do this in tensorflow?
For the sake of completeness, I planed to use
def create_model():
model = keras.Sequential()
model.add(MobileNetV2(input_shape=(224, 224, 3), include_top=False))
model.trainable = True
model.add(layers.GlobalAveragePooling2D())
model.add(layers.Dense(1024, activation="relu"))
model.add(layers.Dense(1, activation="softmax"))
model.compile(optimizer='adam',
loss=tf.losses.mean_squared_error,
metrics=[tf.metrics.SparseCategoricalAccuracy()])
model.summary()
return model
for training the model. The question is only regarding how to read the training data?

Answer 1

I will answer my own question.
The best way was to write a manual function that reads the labels and images.
Assume that the images are in 'data\Images' and the labels are in a .txt file and the labels are in a .txt file at 'data\train_test_files\All_labels.txt'. Then the following two methods will do the job:
def loadImages(IMG_SIZE):
path = os.path.join(os.getcwd(), 'data\\Images')
training_data=[]
labelMap = getLabelMap()
for img in os.listdir(path):
out_array = np.zeros((350,350, 3), np.float32) #350x350 is the pixel size of the images
try:
img_array = cv2.imread(os.path.join(path, img))
img_array=img_array.astype('float32')
out_array = cv2.normalize(img_array, out_array, 0, 1, cv2.NORM_MINMAX)
out_array = cv2.resize(out_array, (IMG_SIZE, IMG_SIZE)
training_data.append([out_array, float(labelMap[img])])
except Exception as e:
pass
return training_data
def getLabelMap():
map = {}
path = os.getcwd()
path = os.path.join(path, "data\\train_test_files\\All_labels.txt")
f = open(path, "r")
for line in f:
line = line.split() #lines in txt file are of the form 'image_name.jpg 3.2'
map[line[0]] = line[1] #3.2 is the label
f.close()
return map
#call of method:
training_set=[]
training_set = loadImages(244) #I want to have my images resized to 244x244