def resize_if_needed(self, image_name):
image = load_image(image_name)
width, height = image.size
if width == self.image_width and height == self.image_height:
return load_image(image_name)
resized_image = resize_image(image, self.image_width,
self.image_height, self.image_channels,
'half_crop')
return scipy.misc.toimage(resized_image, cmin=0.0, cmax=...)
def get_and_seperate_data_X_y(directory, dimension, modes_convolution=None):
dir_list = glob.glob(directory + '*.jpg')
iterator = 1
dataset_X = []
dataset_y = []
for dir_file in dir_list:
origin_image = image_processor.load_image(dir_file)
noise_image = noise_generator("s&p", origin_image)
noise_image = noise_generator("gauss", noise_image)
#convolution
if modes_convolution == None:
noise_image = image_processor.convolutions_image(
noise_image, modes=["gaussian_blur_5_5", "sharpen"])
else:
noise_image = image_processor.convolutions_image(
noise_image, modes=modes_convolution)
gray_origin_image = cv.cvtColor(origin_image, cv.COLOR_BGR2GRAY)
gray_noise_image = cv.cvtColor(noise_image, cv.COLOR_BGR2GRAY)
num_idx_col = int(len(gray_origin_image) / dimension)
num_idx_row = int(len(gray_origin_image[0]) / dimension)
for idx_col in range(num_idx_col):
for idx_row in range(num_idx_row):
X = gray_noise_image[idx_col * dimension:(idx_col + 1) *
dimension, idx_row *
dimension:(idx_row + 1) * dimension]
y = gray_origin_image[idx_col * dimension:(idx_col + 1) *
dimension, idx_row *
dimension:(idx_row + 1) * dimension]
dataset_X.append(X)
dataset_y.append(y)
print("...image number " + str(iterator) + " has been processing")
iterator += 1
print("***DATASET IMAGE HAS BEEN DONE PROCESSED***")
return [dataset_X, dataset_y]
# load pretrained model here!
model = load_model()
for class_index, img_dir in enumerate(img_dirs):
path_images = os.path.join(root_path, img_dir)
image_files = list(filter(lambda x: ".DS" not in x, os.listdir(path_images)))
num_images = len(image_files)
# index splitting in train, test and val
first_index = int(num_images * 0.8)
second_index = first_index + ((num_images - first_index) // 2)
for data_split_index, img_file in enumerate(image_files):
img_file = os.path.join(path_images, img_file)
# rezise image to 299 and pad with black
img = load_image(img_file)
img = prepare_image(img, 299)
# create onehot vector
one_hot = np.zeros(num_classes)
one_hot[class_index] = 1
feature = feature_extract_image(img, model)
if data_split_index < first_index:
num_train_sample += 1
train_set["sample_" + str(num_train_sample)] = np.asarray([feature, one_hot])
elif data_split_index < second_index:
num_val_sample += 1
val_set["sample_" + str(num_val_sample)] = np.asarray([feature, one_hot])
else:
# -*- coding: utf-8 -*-
"""
Created on Tue Jul 3 20:45:33 2018
@author: Davit
"""
from image_processor import load_image
import pandas as pd
import numpy
import cv2
#%%
img_3d = load_image("dataset/Image/All Gambar Rontgen/03 (3).jpg")
df_3d_b = pd.DataFrame(img_3d[:, :, 0])
df_3d_g = pd.DataFrame(img_3d[:, :, 1])
df_3d_r = pd.DataFrame(img_3d[:, :, 2])
#%%
df_3d_b.to_csv("result_report/Grayscaling/df_3d_b.csv")
df_3d_g.to_csv("result_report/Grayscaling/df_3d_g.csv")
df_3d_r.to_csv("result_report/Grayscaling/df_3d_r.csv")
#%%
#grayscale counting