def generate_data(path):
corrupted = ['indooPool_Inside_gif.jpg']
# file with 10 folders
file_list = get_path_list(path)
for i in range(0 , len(file_list) , 1): # i is a label
# specific file of the 10
class_name = file_list[i]
class_path = os.path.join(path , class_name)
# images paths in 1 folder of the 10 folders
class_path_list = get_path_list(class_path)
# iterate throw the images of 1 class
for j in range(0 , len(class_path_list) , 1):
img_name = class_path_list[j]
final_path = os.path.join(class_path ,img_name)
# corrupted images
if class_path_list[j] in corrupted:
continue
img = cv2.imread(final_path, -1)
# bad image
if check_image(img) == 0:
continue
# preprocess the img , then agument the img , then save it in the new file
pimg = preprocessing(img)
img_l = Agumentation(pimg)
label = i
# save the image
save_new_imgs(img_l , img_name , class_name)
return
def test_img(img,split_comb,flag):
alpha = 0.5
red = (0,0,255)
green = (0,128,0)
yellow = (0,255,255)
orange = (0,255,165)
image = cv2.imread(img)
grey = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
p_image = preprocessing(grey)
n_row,n_col = split_comb
instance = splitter(p_image,n_row,n_col)
new_im = []
for i in range(len(flag)):
backtorgb = cv2.cvtColor(instance[i],cv2.COLOR_GRAY2RGB)
overlay = backtorgb.copy()
output = backtorgb.copy()
if flag[i] <= 1:
cv2.rectangle(overlay, (0, 0), (174, 64), red, -1)
elif flag[i] == 2:
cv2.rectangle(overlay, (0, 0), (174, 64), orange, -1)
elif flag[i] == 3:
cv2.rectangle(overlay, (0, 0), (174, 64), yellow, -1)
else:
cv2.rectangle(overlay, (0, 0), (174, 64), green, -1)
cv2.addWeighted(overlay, alpha, output, 1 - alpha, 0, output)
outputImage = cv2.copyMakeBorder(output,1,1,1,1,cv2.BORDER_CONSTANT,value=(0,0,0))
new_im.append(outputImage)
out_im = np.concatenate([np.concatenate(new_im[(i*n_col):((i+1)*n_col)],axis=1) for i in range(n_row)],axis=0)
return(out_im)
def feature(file, split=None):
image = cv2.imread(file)
grey = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
p_image = preprocessing(grey)
if split != None:
instance = splitter(p_image, split[0], split[1])
features = pd.DataFrame([
all_features(np.array(inst), np.array(inst)) for inst in instance
])
features.index = [
os.path.basename(file).replace('.png', '') + '_' + str(k)
for k in range(1, split[0] * split[1] + 1)
]
else:
features = pd.DataFrame(
[all_features(np.array(p_image), np.array(p_image))])
features.index = [os.path.basename(file).replace('.png', '')]
return (features)
def multi_user_add(src_path, out_path):
#Reading all images from source Path
cust = {}
for i in src_path:
j = os.path.basename(i).rsplit('_', 1)[0]
if j not in cust:
cust[j] = []
cust[j].append(i)
cnt = 0
replaced = 0
drop = 0
for cust_id in cust:
if len(cust[cust_id]) < 4:
drop = drop + 1
continue
if os.path.isdir(os.path.join(out_path, cust_id)):
replaced = replaced + 1
shutil.rmtree(os.path.join(out_path, cust_id))
os.mkdir(os.path.join(out_path, cust_id))
os.mkdir(os.path.join(out_path, cust_id, "original"))
os.mkdir(os.path.join(out_path, cust_id, "processed"))
for i in range(len(cust[cust_id])):
shutil.copy(
cust[cust_id][i],
os.path.join(out_path, cust_id, "original",
str(i + 1).zfill(2) + ".png"))
image = cv2.imread(cust[cust_id][i])
grey = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
p_image = preprocessing(grey)
cv2.imwrite(
os.path.join(out_path, cust_id, "processed",
str(i + 1).zfill(2) + ".png"), p_image)
train_files(out_path, cust_id)
cnt = cnt + 1
return ([cnt - replaced, replaced, drop])
def main():
print("test main function ")
obj = preprocessing(test_string)
input = obj.lowercasing(input_string)
print("lower case word is {}".format(input))
stop_input = obj.stopwordemoval(input)
print("stop result is {}".format(stop_input))
utli_obj = utility()
print(utli_obj.get_path(folder_path))
path = utli_obj.get_path(folder_path)
print(utli_obj.get_file_list(path))
class_folder = utli_obj.get_file_list(path)
count = 0
class_count = 0
data = []
for folder in utli_obj.get_file_list(path):
print("folder name's are {} \n".format(folder))
#print("list of file is {}".format(utli_obj.get_file_list(os.path.join(path,folder))))
for file in utli_obj.get_file_list(os.path.join(path, folder)):
print("print fil ein folder is {}".format(file))
#utility=utility()
file_f = utli_obj.file_open(
os.path.join(os.path.join(path, folder), file))
data.append(utli_obj.read_file(file_f))
print("data in file is {}".format(data))
count += 1
if count == 5:
break
#class_folder[class_count]=folder
utli_obj.write_csv_file(data, class_folder)
def single_user_add(img_paths, out_path, cust_id=""):
if not (isinstance(img_paths, list) or isinstance(img_paths, tuple)):
return ("Error")
if len(img_paths) < 4:
return ("Select atleast 4 Signatures")
if cust_id == "":
cust_id = "new_user"
i = 1
while cust_id in os.listdir(out_path):
cust_id = "new_user_" + str(i).zfill(2)
i = i + 1
del i
if os.path.isdir(os.path.join(out_path, cust_id)):
shutil.rmtree(os.path.join(out_path, cust_id))
os.mkdir(os.path.join(out_path, cust_id))
os.mkdir(os.path.join(out_path, cust_id, "original"))
os.mkdir(os.path.join(out_path, cust_id, "processed"))
for i in range(len(img_paths)):
shutil.copy(
img_paths[i],
os.path.join(out_path, cust_id, "original",
str(i + 1).zfill(2) + ".png"))
image = cv2.imread(img_paths[i])
grey = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
p_image = preprocessing(grey)
cv2.imwrite(
os.path.join(out_path, cust_id, "processed",
str(i + 1).zfill(2) + ".png"), p_image)
train_files(out_path, cust_id)
return (cust_id)
def batch_preproc(test):
dfs_lst = []
for i in tqdm(range(0, len(test))):
test_file = test[i]
# Verify that test file is in the directory
if test_file in files:
print('File is in there')
else:
print('Trouble finding ' + str(test_file))
continue
test_path = os.path.join(abfs, test_file)
try:
df_p = preprocessing(test_path, plot=False)
except:
print('Problem with ' + test_file)
continue
df_p['Original File'] = test_file
dfs_lst.append(df_p)
dff = pd.concat(dfs_lst)
return dff
series_mean = series_array.mean(axis=1).reshape(-1, 1)
series_array = series_array - series_mean
series_array = series_array.reshape(
(series_array.shape[0], series_array.shape[1], 1))
return series_array, series_mean
def transform_series_decode(series_array, encode_series_mean):
series_array = np.log1p(np.nan_to_num(series_array)) # filling NaN with 0
series_array = series_array - encode_series_mean
series_array = series_array.reshape(
(series_array.shape[0], series_array.shape[1], 1))
return series_array
series_array, data_start_date, data_end_date, train_pred_start, train_pred_end, train_enc_start, \
train_enc_end, val_enc_start, val_enc_end, date_to_index = preprocessing()
#### Build the model ####
latent_dim = 50 # LSTM hidden units
dropout = .20
# Define an input series and encode it with an LSTM.
encoder_inputs = Input(shape=(None, 1))
encoder = LSTM(latent_dim, dropout=dropout, return_state=True)
encoder_outputs, state_h, state_c = encoder(encoder_inputs)
# We discard `encoder_outputs` and only keep the final states. These represent the "context"
# vector that we use as the basis for decoding.
encoder_states = [state_h, state_c]
# Set up the decoder, using `encoder_states` as initial state.
import os
from utility import utility
import pandas as pd
from model import Sentimentmodel
from sklearn.model_selection import train_test_split
import numpy as np
import matplotlib.pyplot as plt
from sklearn.preprocessing import OneHotEncoder
folder_path="tokens"
class_folder=[]
batchsize=8
epoch=2
vsplit=0.2
verbose=0.2
preprocessing_obj= preprocessing()
ytrain=[1,2,3,4,5,1,2,0,2,2]
def pre_process(input):
#X_t=preprocessing_obj.tokenization(input)
#X_train=preprocessing_obj.stopwordemoval(input)
#X_t=preprocessing_obj.lowercasing(X_train)
#print("X_train is {}".format(X_t))
pre_process_val=preprocessing_obj.bagofwords(input)
print("Bag of words for this {}".format(len(pre_process_val)))
f = open("bagofword.txt", "a")
f.write((str(pre_process_val[:100])))
f.close()
return pre_process_val
