Dipeshpal

import cv2


def make_blur(img):
    # image = cv2.imread(img)
    ksize = (10, 10)
    blur = cv2.blur(img, ksize)
    return blur

Dipeshpal

50/50 [==============================] - 109s 2s/step - loss: 0.2327 - acc: 0.4357 - val_loss: 0.2934 - val_acc: 0.4747 - lr: 1.0000e-04
Epoch 80/200
50/50 [==============================] - ETA: 0s - loss: 0.2390 - acc: 0.4316
Epoch 00080: val_loss did not improve from 0.28716
Training: epoch 80 ends at 20:04:06.414791

50/50 [==============================] - 114s 2s/step - loss: 0.2390 - acc: 0.4316 - val_loss: 0.3010 - val_acc: 0.5238 - lr: 1.0000e-04
Epoch 81/200
50/50 [==============================] - ETA: 0s - loss: 0.2358 - acc: 0.4557
Epoch 00081: val_loss did not improve from 0.28716

Dipeshpal

Model: "model_3"
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_4 (InputLayer)            [(None, 512, 512, 3) 0                                            
__________________________________________________________________________________________________
conv2d_54 (Conv2D)              (None, 512, 512, 16) 448         input_4[0][0]                    
__________________________________________________________________________________________________
dropout_27 (Dropout)            (None, 512, 512, 16) 0           conv2d_54[0][0]                  
__________________________________________________________________________________________________

Dipeshpal

# Trainable params: 1,941,139
class automaticmaplabelling():
    def __init__(self, modelPath, width=512,height=512,channels=3):
        self.modelPath=modelPath
        self.IMG_WIDTH=width
        self.IMG_HEIGHT=height
        self.IMG_CHANNELS=channels
        self.model = self.U_net()

Dipeshpal

model_name = "models/"+"Unet_purple_background.h5"

modelcheckpoint = ModelCheckpoint(model_name,
                                  monitor='val_loss',
                                  mode='auto',
                                  verbose=1,
                                  save_best_only=True)

lr_callback = ReduceLROnPlateau(min_lr=0.000001)

Dipeshpal

model = get_model()
model.summary()

Dipeshpal

import datetime

class MyCustomCallback(tf.keras.callbacks.Callback):
    def on_train_begin(self, epoch, logs={}):

        res_dir = "intermediate_results_purple_background"

        try:
            os.makedirs(res_dir)
        except:

Dipeshpal

def get_model():
    in1 = Input(shape=(IMG_HEIGHT, IMG_WIDTH, 3 ))

    conv1 = Conv2D(32, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(in1)
    conv1 = Dropout(0.2)(conv1)
    conv1 = Conv2D(32, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(conv1)
    pool1 = MaxPooling2D((2, 2))(conv1)

    conv2 = Conv2D(64, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(pool1)
    conv2 = Dropout(0.2)(conv2)

Dipeshpal

cv2_imshow(x[0] * 255.)
cv2_imshow(y['seg'][0] * 255.)

Dipeshpal

def keras_generator_train_val_test(batch_size, choice="train"):

    if choice == "train":
        X = X_train
        y = y_train
    elif choice == "val":
        X = X_val
        y = y_val
    elif choice == "test":
        X = X_test