|
| int | example4 - train.all_cropped_num = len(os.listdir(train_cropped_images_path))//3 |
| |
| | example4 - train.callbacks |
| |
| int | example4 - train.channels = 2 |
| |
| | example4 - train.col |
| |
| | example4 - train.col_end = col+cropped_w |
| |
| | example4 - train.compiled_model = model |
| |
| list | example4 - train.config_list = [(noisy_images, False), (pure_images, False), (ir_images, True)] |
| |
| | example4 - train.conv1 = Conv2D(64, 3, activation='relu', padding='same', kernel_initializer='he_normal')(inputs) |
| |
| | example4 - train.conv10 = Conv2D(channels, 1, activation='sigmoid')(conv9) |
| |
| | example4 - train.conv2 = Conv2D(128, 3, activation='relu', padding='same', kernel_initializer='he_normal')(pool1) |
| |
| | example4 - train.conv3 = Conv2D(256, 3, activation='relu', padding='same', kernel_initializer='he_normal')(pool2) |
| |
| | example4 - train.conv4 = Conv2D(512, 3, activation='relu', padding='same', kernel_initializer='he_normal')(pool3) |
| |
| | example4 - train.conv5 = Conv2D(1024, 3, activation='relu', padding='same', kernel_initializer='he_normal')(pool4) |
| |
| | example4 - train.conv6 = Conv2D(512, 3, activation='relu', padding='same', kernel_initializer='he_normal')(merge6) |
| |
| | example4 - train.conv7 = Conv2D(256, 3, activation='relu', padding='same', kernel_initializer='he_normal')(merge7) |
| |
| | example4 - train.conv8 = Conv2D(128, 3, activation='relu', padding='same', kernel_initializer='he_normal')(merge8) |
| |
| | example4 - train.conv9 = Conv2D(64, 3, activation='relu', padding='same', kernel_initializer='he_normal')(merge9) |
| |
| | example4 - train.crop = img.crop((col_i, row_i, col_i + w, row_i + h)) |
| |
| | example4 - train.cropped_h |
| |
| list | example4 - train.cropped_image_offsets = [] |
| |
| list | example4 - train.cropped_images_list = [(cropped_noisy_images, "noisy"), (cropped_pure_images, "pure")] |
| |
| list | example4 - train.cropped_ir_images = [f for f in glob.glob(train_cropped_images_path + "**/left*" + IMAGE_EXTENSION, recursive=True)] |
| |
| list | example4 - train.cropped_noisy_images = [f for f in glob.glob(train_cropped_images_path + "**/res*" + IMAGE_EXTENSION, recursive=True)] |
| | convert cropped images to arrays More...
|
| |
| list | example4 - train.cropped_pure_images = [f for f in glob.glob(train_cropped_images_path + "**/gt*" + IMAGE_EXTENSION, recursive=True)] |
| |
| | example4 - train.cropped_w |
| |
| | example4 - train.curr_cropped_images |
| |
| | example4 - train.denoised_col = cropped_w |
| |
| string | example4 - train.denoised_dir = images_path+r"/denoised" |
| |
| | example4 - train.denoised_image = model.predict(sample) |
| |
| | example4 - train.denoised_name = os.path.basename(directory.split('/')[-1]) |
| |
| | example4 - train.denoised_row = cropped_h |
| |
| | example4 - train.drop4 = Dropout(0.5)(conv4) |
| |
| | example4 - train.drop5 = Dropout(0.5)(conv5) |
| |
| | example4 - train.epochs |
| |
| | example4 - train.file_path = os.path.join(train_cropped_images_path, filename) |
| |
| | example4 - train.filelist |
| |
| | example4 - train.first_image = i*images_num_to_process |
| |
| int | example4 - train.frame_num = 0 |
| |
| | example4 - train.gpus = tf.config.experimental.list_physical_devices('GPU') |
| |
| | example4 - train.gray_image = cv2.cvtColor(ii, cv2.COLOR_BGR2GRAY) |
| |
| | example4 - train.h |
| |
| | example4 - train.height |
| |
| | example4 - train.ii = cv2.imread(file) |
| |
| | example4 - train.im_and_ir = images_plt |
| |
| | example4 - train.im_files = [f for f in glob.glob(directory + "**/res*" , recursive=True)] |
| |
| string | example4 - train.IMAGE_EXTENSION = '.png' |
| |
| int | example4 - train.images_num_to_process = 1000 |
| |
| string | example4 - train.images_path = root+r"/images" |
| |
| list | example4 - train.images_plt = [cv2.imread(f, cv2.IMREAD_UNCHANGED) for f in im_files if f.endswith(IMAGE_EXTENSION)] |
| |
| | example4 - train.images_type |
| |
| | example4 - train.img = Image.fromarray(np.array(gray_image).astype("uint16")) |
| |
| | example4 - train.img_height |
| |
| | example4 - train.img_width |
| |
| tuple | example4 - train.input_size = (img_width, img_height, channels) |
| |
| | example4 - train.inputs = Input(input_size) |
| |
| tuple | example4 - train.ir_config = (ir_images, ir_total_cropped_images, True, {}) |
| | SPLIT IMAGES ##################. More...
|
| |
| string | example4 - train.ir_cropped_images_file = test_cropped_images_path+r'/' |
| |
| | example4 - train.ir_im_files = [f for f in glob.glob(ir_cropped_images_file + "**/left*" , recursive=True)] |
| |
| list | example4 - train.ir_images = [f for f in glob.glob(train_images + "**/left*" + IMAGE_EXTENSION, recursive=True)] |
| |
| list | example4 - train.ir_images_plt = [cv2.imread(f, cv2.IMREAD_UNCHANGED) for f in ir_im_files if f.endswith(IMAGE_EXTENSION)] |
| |
| list | example4 - train.ir_total_cropped_images = [0]*len(ir_images) |
| |
| | example4 - train.is_ir |
| |
| | example4 - train.iterations = all_cropped_num//images_num_to_process |
| |
| | example4 - train.limit = first_image+images_num_to_process |
| |
| | example4 - train.log_file = open(name, "w") |
| |
| | example4 - train.logical_gpus = tf.config.experimental.list_logical_devices('GPU') |
| |
| string | example4 - train.logs_path = root+r"/logs" |
| |
| | example4 - train.loss |
| |
| | example4 - train.merge6 = concatenate([drop4, up6], axis=3) |
| |
| | example4 - train.merge7 = concatenate([conv3, up7], axis=3) |
| |
| | example4 - train.merge8 = concatenate([conv2, up8], axis=3) |
| |
| | example4 - train.merge9 = concatenate([conv1, up9], axis=3) |
| |
| | example4 - train.metrics |
| |
| | example4 - train.model = Model(inputs=inputs, outputs=conv10) |
| |
| | example4 - train.model_checkpoint = ModelCheckpoint(models_path + r"/unet_membrane.hdf5", monitor='loss', verbose=1, save_best_only=True) |
| |
| string | example4 - train.model_name = 'DEPTH_' |
| |
| string | example4 - train.models_path = root+r"/models" |
| |
| string | example4 - train.name = logs_path+r'/loss_output_' |
| |
| string | example4 - train.new_test_cropped_images_path = test_cropped_images_path+r'/' |
| |
| tuple | example4 - train.noisy_config = (noisy_images, total_cropped_images, False, origin_files_index_size_path_test) |
| |
| list | example4 - train.noisy_images = [f for f in glob.glob(train_images + "**/res*" + IMAGE_EXTENSION, recursive=True)] |
| |
| | example4 - train.noisy_input_train = img |
| |
| | example4 - train.old_stdout = sys.stdout |
| |
| | example4 - train.optimizer |
| |
| | example4 - train.origin_file_name |
| |
| | example4 - train.origin_files_index_size_path |
| |
| dictionary | example4 - train.origin_files_index_size_path_test = {} |
| |
| string | example4 - train.outfile = denoised_dir+'/' |
| |
| | example4 - train.path = os.path.join(train_cropped_images_path, curr_cropped_images[i]) |
| |
| list | example4 - train.paths = [root, images_path, models_path, logs_path, train_images, train_cropped_images_path] |
| |
| | example4 - train.pool1 = MaxPooling2D(pool_size=(2, 2))(conv1) |
| |
| | example4 - train.pool2 = MaxPooling2D(pool_size=(2, 2))(conv2) |
| |
| | example4 - train.pool3 = MaxPooling2D(pool_size=(2, 2))(conv3) |
| |
| | example4 - train.pool4 = MaxPooling2D(pool_size=(2, 2))(drop4) |
| |
| list | example4 - train.pure_images = [f for f in glob.glob(train_images + "**/gt*" + IMAGE_EXTENSION, recursive=True)] |
| |
| | example4 - train.pure_input_train = img |
| |
| int | example4 - train.rolling_frame_num = 0 |
| |
| string | example4 - train.root = r"./unet_flow" |
| |
| | example4 - train.row |
| |
| | example4 - train.row_end = row+cropped_h |
| |
| | example4 - train.sample = samples[i:i+1] |
| |
| | example4 - train.samples = img |
| |
| string | example4 - train.save_model_name = models_path+'/' |
| |
| | example4 - train.save_to |
| |
| | example4 - train.stdout |
| |
| | example4 - train.steps_per_epoch = len(cropped_noisy_images)//unet_epochs |
| |
| | example4 - train.t1 = time.perf_counter() |
| |
| | example4 - train.t2 = time.perf_counter() |
| |
| string | example4 - train.test_cropped_images_path = images_path+r"/test_cropped" |
| |
| string | example4 - train.test_images = images_path+r"/test" |
| |
| | example4 - train.test_img_height |
| |
| | example4 - train.test_img_width |
| |
| | example4 - train.test_model_name = save_model_name |
| |
| | example4 - train.timestr = time.strftime("%Y%m%d-%H%M%S") |
| |
| list | example4 - train.total_cropped_images = [0]*len(noisy_images) |
| |
| string | example4 - train.train_cropped_images_path = images_path+r"/train_cropped" |
| |
| string | example4 - train.train_images = images_path+r"/train" |
| |
| int | example4 - train.unet_epochs = 1 |
| |
| | example4 - train.up6 = Conv2D(512, 2, activation='relu', padding='same', kernel_initializer='he_normal')(UpSampling2D(size=(2, 2))(drop5)) |
| |
| | example4 - train.up7 = Conv2D(256, 2, activation='relu', padding='same', kernel_initializer='he_normal')(UpSampling2D(size=(2, 2))(conv6)) |
| |
| | example4 - train.up8 |
| |
| | example4 - train.up9 = Conv2D(64, 2, activation='relu', padding='same', kernel_initializer='he_normal')(UpSampling2D(size=(2, 2))(conv8)) |
| |
| | example4 - train.w |
| |
| | example4 - train.whole_image = np.zeros((height, width, channels), dtype="float32") |
| |
| | example4 - train.width |
| |