def resize_data():
files = getFiles(test_dir)
for index, file in enumerate(files):
filename = file.split('/')[-1]
dist_file_path = test_dir + '2' + filename
img = cv2.imread(file, cv2.IMREAD_ANYDEPTH | cv2.IMREAD_ANYCOLOR)
img2 = np.resize(img, (128, 128))
cv2.imwrite(dist_file_path, img2)
def process_train_data(pic_url, img_size, save_url):
files = getFiles(pic_url)
total_data = np.zeros((len(files), img_size * img_size), dtype=np.float32)
for index, path in enumerate(files):
img = cv2.imread(path, cv2.IMREAD_ANYCOLOR | cv2.IMREAD_ANYDEPTH)
total_data[index, :] = img.flatten()
total_data = total_data / 255
np.save(save_url, total_data)
return total_data
def gen_label_file(dir, dist, type='target'):
if type == 'target':
files = getFiles(dir)
val = 1
else:
files = getFiles_jpg(dir)
val = 0
with open(dist, 'w') as label_file:
for file_path in files:
label_file.write(file_path + ',' + str(val))
label_file.write('\n')
def generate_train_label_back_data(pic_url, img_size, save_url):
files = getFiles(pic_url)
total_data = np.zeros((len(files), img_size * img_size), dtype=np.float32)
for index, path in enumerate(files):
img = cv2.imread(path, cv2.IMREAD_ANYCOLOR | cv2.IMREAD_ANYDEPTH)
if img is None:
print path
continue
total_data[index, :] = img.flatten()
total_data /= 255
label_data = np.zeros((len(files), 3), dtype=np.float32)
label_data[:, 1] = 1
save_file = open(save_url, "wb")
pickle.dump([total_data, label_data], save_file)
print total_data.shape
print label_data.shape
def generate_train_data(sample_count):
files_target = getFiles(target_resize_dir)
files_clutter = getFiles_jpg(clutter_dir)
# files = getFiles_jpg(clutter_dir)
target_data = np.zeros([13411, 128 * 128])
clutter_data = np.zeros([1159, 128 * 128])
for index, file in enumerate(files_target):
img = cv2.imread(file, cv2.IMREAD_ANYDEPTH | cv2.IMREAD_ANYCOLOR)
target_data[index, :] = img.flatten()
# cv2.imshow('sar_target', img)
# cv2.waitKey(0)
# print target_data.shape
for index, file in enumerate(files_clutter):
img = cv2.imread(file, cv2.IMREAD_ANYDEPTH | cv2.IMREAD_ANYCOLOR)
clutter_data[index, :] = img.flatten()
# cv2.imshow('sar_target', img)
# cv2.waitKey(0)
# print clutter_data.shape
target_data_index = np.random.permutation(13411)
clutter_data_index = np.random.permutation(1159)
train_data_set = np.zeros([2 * sample_count, 128 * 128])
train_data_set[:sample_count, :] = target_data[
target_data_index[0:sample_count]]
train_data_set[sample_count:, :] = clutter_data[
clutter_data_index[0:sample_count]]
test_data_count = 13411 + 1159 - 2 * sample_count
test_data_set = np.zeros([test_data_count, 128 * 128])
test_target_count = 13411 - sample_count
test_data_set[:test_target_count, :] = target_data[
target_data_index[sample_count:]]
test_data_set[test_target_count:, :] = clutter_data[
clutter_data_index[sample_count:]]
train_labels = np.zeros([2 * sample_count])
train_labels[0:sample_count] = 1
test_labels = np.zeros([test_data_count])
test_labels[:test_target_count] = 1
# print train_data_set.shape, train_labels.shape, test_data_set.shape, test_labels.shape
return train_data_set, train_labels, test_data_set, test_labels
import numpy as np
import tensorflow as tf
from theano_rbm.data_process import getFiles
import os
import cv2
if __name__ == '__main__':
PATCH_SIZE = 25
target_test = '/home/aurora/hdd/workspace/data/MSTAR_data_liang_processed/target_chips_128x128_normalized_wei_counter/patch_size_25_new/target_test/'
shadow_test = '/home/aurora/hdd/workspace/data/MSTAR_data_liang_processed/target_chips_128x128_normalized_wei_counter/patch_size_25_new/shadow_test/'
bg_test = '/home/aurora/hdd/workspace/data/MSTAR_data_liang_processed/target_chips_128x128_normalized_wei_counter/patch_size_25_new/bg_test/'
files = getFiles(target_test)
filelist = []
FileNames = os.listdir(target_test)
if len(FileNames) > 0:
for fn in FileNames:
if 'HB03848' in fn:
fullfilename = os.path.join(target_test, fn)
filelist.append(fullfilename)
total_data = np.zeros((len(filelist), PATCH_SIZE * PATCH_SIZE),
dtype=np.float32)
for index, path in enumerate(filelist):
img = cv2.imread(path, cv2.IMREAD_ANYCOLOR | cv2.IMREAD_ANYDEPTH)
if img is None:
print path
continue
total_data[index, :] = img.flatten()
print index, path
total_data /= 255
# img_back = total_data[-1].reshape(25, 25)
cv2.imwrite(save_url+img_name, img_all)
return img_all
def get_file_name(url):
filename = url.split('/')[-1][:-4]
return filename
if __name__ == "__main__":
save_url = '/home/aurora/hdd/workspace/data/MSTAR_data_liang_processed/target_chips_128x128_normalized_wei_counter/'
save_test_url = '/home/aurora/hdd/workspace/data/MSTAR_data_liang_processed/target_chips_128x128_normalized_wei_counter/patch_size_5/test_counter/'
img_url = '/home/aurora/hdd/workspace/data/MSTAR_data_liang_processed/target_chips_128x128_normalized_wei/'
xml_url = '/home/aurora/hdd/workspace/PycharmProjects/sar_edge_detection/tf_sda/annoation_xml'
annotation_test_xml = '/home/aurora/hdd/workspace/PycharmProjects/sar_edge_detection/tf_sda/annoation_test_xml/'
annoation_xml_list = getFiles(annotation_test_xml)
print annoation_xml_list
# file_name_list = [get_file_name(filename) for filename in annoation_xml_list]
for annoation_name in annoation_xml_list:
filename = get_file_name(annoation_name)
img_target = cv2.imread(img_url+filename+'.jpg', cv2.IMREAD_ANYDEPTH | cv2.IMREAD_ANYCOLOR)
img_back = img_target.copy()
handler = get_counter(annoation_name)
img_target = save_target(handler, img_target, save_test_url, filename+'_target.jpg')
img_back = save_back(handler, img_back, save_test_url, filename+'_back.jpg')
img_all = save_all(img_target, img_back, save_test_url, filename+'_all.jpg')
# # remove duplicate files
# annoation_xml_list = getFiles_name(xml_url)
# annoation_test_list = getFiles_name(annotation_test_xml)