def __getitem__(self, idx):
image_root = self.train_image_file_paths[idx]
image_name = image_root.split('/')[-1]
image = Image.open(image_root)
#print(image)
fix_size = (160, 60)
image = image.resize(fix_size)
# print(image_name)
if self.transform is not None:
image = self.transform(image)
# print(image_name)
if ('_' in image_name):
label = ohe.encode(image_name.split('_')[0].upper())
else:
label = ohe.encode(image_name.split('.')[0].upper())
return image, label, image_name
def __getitem__(self, idx):
image_root = self.train_image_file_paths[idx]
image_name = image_root.split(os.path.sep)[-1]
image = Image.open(image_root)
if self.transform is not None:
image = self.transform(image)
label = ohe.encode(image_name.split('_')[0])
return image, label
def __getitem__(self, idx):
image_root = self.train_image_file_paths[idx]
image_name = image_root.split(os.path.sep)[-1]
image = Image.open(image_root)
if self.transform is not None:
image = self.transform(image)
label = ohe.encode(image_name.split('_')[0]) # 为了方便，在生成图片的时候，图片文件的命名格式 "4个数字或者数字_时间戳.PNG", 4个字母或者即是图片的验证码的值，字母大写,同时对该值做 one-hot 处理
return image, label
def __getitem__(self, idx):
image_root = self.train_image_file_paths[idx]
image_name = image_root.split(os.path.sep)[-1]
image = Image.open(image_root)
if self.transform is not None:
image = self.transform(image)
label = ohe.encode(
image_name.split('_')[0]
) # For convenience, we name the img file as "4letters_timestamp.PNG", and then one-hot encode it
return image, label
def _preload(self):
"""preload dataset to momory"""
self.labels = []
self.images = []
for image_fn in self.filenames:
image = cv2.imread(image_fn)
self.images.append(image.copy())
label_string = image_fn[-8:-4]
label = ohe.encode(label_string)
self.labels.append(label)
def __getitem__(self, idx):
image_root = self.train_image_file_paths[idx]
image_name = image_root.split(os.path.sep)[-1]
image = Image.open(image_root)
image = image.resize((self.iw, self.ih))
# print("width" + str(image.width) + " height" + str(image.height))
if self.transform is not None:
image = self.transform(image)
x1 = image[np.newaxis, :]
name_ = image_name[0:4]
label = ohe.encode(
name_
) # 为了方便，在生成图片的时候，图片文件的命名格式 "4个数字_时间戳.PNG", 4个数字即是图片的验证码的值,同时对该值做 one-hot 处理
return image, label
def __getitem__(self, idx):
"""Get a sample from the dataset"""
if self.images:
# if preload
image = self.images[idx]
label = self.labels[idx]
else:
image_name = self.filenames[idx]
image = cv2.imread(image_name, 0)
label_string = image_name[-8:-4]
label = ohe.encode(label_string)
# if we have transform functions
if self.transform:
image = self.transform(image)
return image, label
# generate nouns
#noun_file_main = main_file + str(i) + 'n' + '.txt'
#noun_eval_file = eval_file + 'n' + '.txt'
## keywords extracted from the main file will normalize all the values
#normalizing_val = extract_keywords(main_file+str(i)+'.txt',noun_file_main)
#print(normalizing_val)
#if normalizing_val>val:
# val = normalizing_val
# temp = i
# print(normalizing_val)
#print(val,temp)
## keywords extracted from the eval file will be stored for clustering
#count_eval = extract_keywords(eval_file+str(i)+'.txt',eval_file+str(i)+'n'+'.txt')
#if count_eval <10:
#os.remove(eval_file+str(i)+'.txt')
#os.remove(eval_file+str(i)+'n'+'.txt')
#print("kam",count_eval)
#print(eval_file+str(i)+'.txt')
# one hot encoding
encode(main_file + 'n' + '.txt', eval_file + str(i) + '.txt', 0)
#print(x,main_score)
# store the results in a file to be used by clustering
# value to be store is main_score and count_eval
#results_file = '/home/manan/Desktop/Research/Learning-Perspectives/results/scores.txt'
#found = count_keywords(eval_file + str(i)+'n'+'.txt')
#write_score(x,main_score,found,results_file)