Create utils file
Browse files
utils.py
ADDED
|
@@ -0,0 +1,230 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
from google.cloud import vision
|
| 3 |
+
import re
|
| 4 |
+
import torch
|
| 5 |
+
import torchvision
|
| 6 |
+
import numpy as np
|
| 7 |
+
from PIL import Image
|
| 8 |
+
import albumentations as A
|
| 9 |
+
from albumentations.pytorch import ToTensorV2
|
| 10 |
+
import tempfile
|
| 11 |
+
import json
|
| 12 |
+
|
| 13 |
+
def getcredentials():
|
| 14 |
+
secret_key_credential = os.getenv("secret_key")
|
| 15 |
+
|
| 16 |
+
with tempfile.NamedTemporaryFile(mode='w+', delete= True, suffix=".json") as temp_file:
|
| 17 |
+
temp_file.write(json.dumps(secret_key_credential))
|
| 18 |
+
|
| 19 |
+
tempfile_name = temp_file.name
|
| 20 |
+
|
| 21 |
+
return tempfile_name
|
| 22 |
+
|
| 23 |
+
|
| 24 |
+
os.environ['GOOGLE_APPLICATION_CREDENTIALS'] = getcredentials()
|
| 25 |
+
|
| 26 |
+
##
|
| 27 |
+
def info_new_cni(donnees):
|
| 28 |
+
##
|
| 29 |
+
informations = {}
|
| 30 |
+
|
| 31 |
+
# Utilisation d'expressions régulières pour extraire les informations spécifiques
|
| 32 |
+
numero_carte = re.search(r'n° (C\d+)', ' '.join(donnees))
|
| 33 |
+
#prenom_nom = re.search(r'Prénom\(s\)\s+(.*?)\s+Nom\s+(.*?)\s+Signature', ' '.join(donnees))
|
| 34 |
+
nom = re.search(r'Nom\s+(.*?)\s', ' '.join(donnees))
|
| 35 |
+
prenom = re.search(r'Prénom\(s\)\s+(.*?)\s+Nom\s+(.*?)', ' '.join(donnees))
|
| 36 |
+
date_naissance = re.search(r'Date de Naissance\s+(.*?)+(\d{2}/\d{2}/\d{4})', ' '.join(donnees))
|
| 37 |
+
lieu_naissance = re.search(r'Lieu de Naissance\s+(.*?)\s', ' '.join(donnees))
|
| 38 |
+
taille = re.search(r'Sexe Taille\s+(.*?)+(\d+,\d+)', ' '.join(donnees))
|
| 39 |
+
nationalite = re.search(r'Nationalité\s+(.*?)\s+\d+', ' '.join(donnees))
|
| 40 |
+
date_expiration = re.search(r'Date d\'expiration\s+(\d+/\d+/\d+)', ' '.join(donnees))
|
| 41 |
+
sexe = re.search(r'Date de Naissance\s+(.*?)+(\d{2}/\d{2}/\d{4})+(.*)', ' '.join(donnees))
|
| 42 |
+
|
| 43 |
+
# Stockage des informations extraites dans un dictionnaire
|
| 44 |
+
if numero_carte:
|
| 45 |
+
informations['Numéro de carte'] = numero_carte.group(1)
|
| 46 |
+
if nom :
|
| 47 |
+
informations['Nom'] = nom.group(1)
|
| 48 |
+
|
| 49 |
+
if prenom:
|
| 50 |
+
informations['Prénom'] = prenom.group(1)
|
| 51 |
+
|
| 52 |
+
if date_naissance:
|
| 53 |
+
informations['Date de Naissance'] = date_naissance.group(2)
|
| 54 |
+
if lieu_naissance:
|
| 55 |
+
informations['Lieu de Naissance'] = lieu_naissance.group(1)
|
| 56 |
+
if taille:
|
| 57 |
+
informations['Taille'] = taille.group(2)
|
| 58 |
+
if nationalite:
|
| 59 |
+
informations['Nationalité'] = nationalite.group(1)
|
| 60 |
+
if date_expiration:
|
| 61 |
+
informations['Date d\'expiration'] = date_expiration.group(1)
|
| 62 |
+
if sexe :
|
| 63 |
+
informations['sexe'] = sexe.group(3)[:2]
|
| 64 |
+
|
| 65 |
+
return informations
|
| 66 |
+
|
| 67 |
+
##
|
| 68 |
+
|
| 69 |
+
def info_ancien_cni(infos):
|
| 70 |
+
""" Extract information in row data of ocr"""
|
| 71 |
+
|
| 72 |
+
informations = {}
|
| 73 |
+
|
| 74 |
+
immatriculation_patern = r'Immatriculation:\s+(C \d{4} \d{4} \d{2})'
|
| 75 |
+
immatriculation = re.search(immatriculation_patern, ''.join(infos))
|
| 76 |
+
nom = infos[4]
|
| 77 |
+
prenom_pattern = r'Nom\n(.*?)\n'
|
| 78 |
+
prenom = re.search(prenom_pattern, '\n'.join(infos))
|
| 79 |
+
sexe_pattern = r'Prénoms\n(.*?)\n'
|
| 80 |
+
sexe = re.search(sexe_pattern, '\n'.join(infos))
|
| 81 |
+
taille_pattern = r'Sexe\n(.*?)\n'
|
| 82 |
+
taille = re.search(taille_pattern, '\n'.join(infos))
|
| 83 |
+
date_naiss_pattern = r'Taille\s+(.*?)+(\d+/\d+/\d+)' # r'Taille (m)\n(.*?)\n'
|
| 84 |
+
date_naissance = re.search(date_naiss_pattern, ' '.join(infos))
|
| 85 |
+
lieu_pattern = r'Date de Naissance\n(.*?)\n'
|
| 86 |
+
lieu_naissance = re.search(lieu_pattern, '\n'.join(infos))
|
| 87 |
+
valide_pattern = r'Valide jusqu\'au+(.*?)+(\d+/\d+/\d+)'
|
| 88 |
+
validite = re.search(valide_pattern, ' '.join(infos))
|
| 89 |
+
|
| 90 |
+
# Stockage des informations extraites dans un dictionnaire
|
| 91 |
+
if immatriculation:
|
| 92 |
+
informations['Immatriculation'] = immatriculation.group(1)
|
| 93 |
+
if nom :
|
| 94 |
+
informations['Nom'] = infos[4]
|
| 95 |
+
|
| 96 |
+
if prenom:
|
| 97 |
+
informations['Prénom'] = prenom.group(1)
|
| 98 |
+
|
| 99 |
+
if date_naissance:
|
| 100 |
+
informations['Date de Naissance'] = date_naissance.group(2)
|
| 101 |
+
if lieu_naissance:
|
| 102 |
+
informations['Lieu de Naissance'] = lieu_naissance.group(1)
|
| 103 |
+
if taille:
|
| 104 |
+
informations['Taille'] = taille.group(1)
|
| 105 |
+
|
| 106 |
+
if validite:
|
| 107 |
+
informations['Date d\'expiration'] = validite.group(2)
|
| 108 |
+
if sexe :
|
| 109 |
+
informations['sexe'] = sexe.group(1)
|
| 110 |
+
|
| 111 |
+
return informations
|
| 112 |
+
|
| 113 |
+
##
|
| 114 |
+
def filtrer_elements(liste):
|
| 115 |
+
elements_filtres = []
|
| 116 |
+
for element in liste:
|
| 117 |
+
if element not in ['\r',"RÉPUBLIQUE DE CÔTE D'IVOIRE", "MINISTÈRE DES TRANSPORTS", "PERMIS DE CONDUIRE"]:
|
| 118 |
+
elements_filtres.append(element)
|
| 119 |
+
return elements_filtres
|
| 120 |
+
|
| 121 |
+
def permis_de_conduite(donnees):
|
| 122 |
+
""" Extraire les information de permis de conduire"""
|
| 123 |
+
|
| 124 |
+
informations = {}
|
| 125 |
+
|
| 126 |
+
infos = filtrer_elements(donnees)
|
| 127 |
+
|
| 128 |
+
nom_pattern = r'Nom\n(.*?)\n'
|
| 129 |
+
nom = re.search(nom_pattern, '\n'.join(infos))
|
| 130 |
+
prenom_pattern = r'Prénoms\n(.*?)\n'
|
| 131 |
+
prenom = re.search(prenom_pattern, '\n'.join(infos))
|
| 132 |
+
date_lieu_naissance_patern = r'Date et lieu de naissance\n(.*?)\n'
|
| 133 |
+
date_lieu_naissance = re.search(date_lieu_naissance_patern, '\n'.join(infos))
|
| 134 |
+
date_lieu_delivrance_patern = r'Date et lieu de délivrance\n(.*?)\n'
|
| 135 |
+
date_lieu_delivrance = re.search(date_lieu_delivrance_patern, '\n'.join(infos))
|
| 136 |
+
numero_pattern = r'Numéro du permis de conduire\n(.*?)\n'
|
| 137 |
+
numero = re.search(numero_pattern, '\n'.join(infos))
|
| 138 |
+
restriction_pattern = r'Restriction\(s\)\s+(.*?)+(.*)'
|
| 139 |
+
restriction = re.search(restriction_pattern, ' '.join(infos))
|
| 140 |
+
|
| 141 |
+
# Stockage des informations extraites dans un dictionnaire
|
| 142 |
+
if nom:
|
| 143 |
+
informations['Nom'] = nom.group(1)
|
| 144 |
+
|
| 145 |
+
if prenom :
|
| 146 |
+
informations['Prenoms'] = prenom.group(1)
|
| 147 |
+
if date_lieu_naissance :
|
| 148 |
+
informations['Date_et_lieu_de_naissance'] = date_lieu_naissance.group(1)
|
| 149 |
+
if date_lieu_naissance :
|
| 150 |
+
informations['Date_et_lieu_de_délivrance'] = date_lieu_delivrance.group(1)
|
| 151 |
+
|
| 152 |
+
informations['Categorie'] = infos[0]
|
| 153 |
+
if numero:
|
| 154 |
+
informations['Numéro_du_permis_de_conduire'] = numero.group(1)
|
| 155 |
+
|
| 156 |
+
if restriction:
|
| 157 |
+
informations['Restriction(s)'] = restriction.group(2)
|
| 158 |
+
|
| 159 |
+
return informations
|
| 160 |
+
|
| 161 |
+
|
| 162 |
+
# Fonction pour extraire les informations individuelles
|
| 163 |
+
def extraire_informations_carte(path, type_de_piece=1):
|
| 164 |
+
""" Detect text in identity card"""
|
| 165 |
+
|
| 166 |
+
client = vision.ImageAnnotatorClient()
|
| 167 |
+
|
| 168 |
+
with open(path,'rb') as image_file:
|
| 169 |
+
content = image_file.read()
|
| 170 |
+
|
| 171 |
+
image = vision.Image(content = content)
|
| 172 |
+
|
| 173 |
+
# for non dense text
|
| 174 |
+
#response = client.text_detection(image=image)
|
| 175 |
+
#for dense text
|
| 176 |
+
response = client.document_text_detection(image = image)
|
| 177 |
+
texts = response.text_annotations
|
| 178 |
+
ocr_texts = []
|
| 179 |
+
|
| 180 |
+
for text in texts:
|
| 181 |
+
ocr_texts.append(f"\r\n{text.description}")
|
| 182 |
+
|
| 183 |
+
if response.error.message :
|
| 184 |
+
raise Exception("{}\n For more informations check : https://cloud.google.com/apis/design/errors".format(response.error.message))
|
| 185 |
+
|
| 186 |
+
donnees = ocr_texts[0].split('\n')
|
| 187 |
+
|
| 188 |
+
if type_de_piece ==1:
|
| 189 |
+
return info_new_cni(donnees)
|
| 190 |
+
elif type_de_piece == 2:
|
| 191 |
+
return info_ancien_cni(donnees)
|
| 192 |
+
elif type_de_piece == 3:
|
| 193 |
+
return permis_de_conduite(donnees)
|
| 194 |
+
else :
|
| 195 |
+
return "Le traitement de ce type de document n'est pas encore pris en charge"
|
| 196 |
+
|
| 197 |
+
def load_checkpoint(path):
|
| 198 |
+
print('--> Loading checkpoint')
|
| 199 |
+
return torch.load(path,map_location=torch.device('cpu'))
|
| 200 |
+
|
| 201 |
+
def make_prediction(image_path):
|
| 202 |
+
|
| 203 |
+
# define the using of GPU or CPU et background training
|
| 204 |
+
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
| 205 |
+
## load model
|
| 206 |
+
model = load_checkpoint("data/model.pth")
|
| 207 |
+
## transformation
|
| 208 |
+
test_transforms = A.Compose([
|
| 209 |
+
A.Resize(height=224, width=224, always_apply=True),
|
| 210 |
+
A.Normalize(always_apply=True),
|
| 211 |
+
ToTensorV2(always_apply=True),])
|
| 212 |
+
|
| 213 |
+
## read the image
|
| 214 |
+
image = np.array(Image.open(image_path).convert('RGB'))
|
| 215 |
+
transformed = test_transforms(image= image)
|
| 216 |
+
image_transformed = transformed["image"]
|
| 217 |
+
image_transformed = image_transformed.unsqueeze(0)
|
| 218 |
+
image_transformed = image_transformed.to(device)
|
| 219 |
+
|
| 220 |
+
model.eval()
|
| 221 |
+
with torch.set_grad_enabled(False):
|
| 222 |
+
output = model(image_transformed)
|
| 223 |
+
|
| 224 |
+
# Post-process predictions
|
| 225 |
+
probabilities = torch.nn.functional.softmax(output[0], dim=0)
|
| 226 |
+
predicted_class = torch.argmax(probabilities).item()
|
| 227 |
+
proba = float(max(probabilities))
|
| 228 |
+
|
| 229 |
+
|
| 230 |
+
return proba, predicted_class
|