tensorflow.python.framework.errors_impl.FailedPreconditionError: Error while reading resource variable predictions/kernel from Container: localhost. This could mean that the variable was uninitialized. Not found: Container localhost does not exist. (Could not find resource: localhost/predictions/kernel)
[[{{node predictions/MatMul/ReadVariableOp}}]]
这是我的代码：
import pandas as pd
import os
import json
import numpy as np
from keras.applications.imagenet_utils import preprocess_input, decode_predictions
from keras.preprocessing.image import img_to_array, load_img #ImageDataGenerator, array_to_img
from keras.models import load_model #Sequential
import base64
import tensorflow as tf
global graph,model
graph = tf.get_default_graph()
import ssl
try:
import httplib # Python 2
except:
import http.client as httplib # Python 3
path = os.path.dirname(os.path.abspath(__file__))
path = path[0:-4]
modelpath = os.path.join(path, r'static/models')
#Loading models # Slower
VGG16_model = load_model(str(modelpath) + '/vgg16.h5')
print ("VGG16_model loaded")
DamgeDetection_model = load_model(str(modelpath) + '/car_damage-model_retrained.h5')
print ("DamgeDetection_model loaded")
location_model = load_model(str(modelpath) + '/carLocationWithfrontAndBack3400Images.h5')
print ("Location model loaded")
severity_model = load_model(str(modelpath)+ '/car_damage_severity.h5')
print ("Severity model loaded")
def prepare_img_224(img_path):
img = load_img(img_path, target_size=(224, 224))
x = img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
return x
def prepare_img_128(img_path):
img = load_img(img_path, target_size=(128 , 128))
x = img_to_array(img)
x = x.reshape((1,) + x.shape)/64
return x
#validating provided image is car
def car_categories_gate(img_224, model):
print ("Validating that this is a picture of your car...")
with graph.as_default():
out = model.predict(img_224)
carList =['sports_car', 'minivan', 'grille', 'convertible', 'racer', 'beach_wagon', 'jeep', 'pickup', 'minibus',
'limousine', 'golfcart', 'cab', 'car_wheel', 'ambulance', 'police_van', 'Model_T',
'tow_truck', 'trailer_truck', 'coffeepot', 'car_mirror', 'snowmobile','chain_saw', 'snowplow',
'fire_engine', 'recreational_vehicle', 'waffle_iron', 'harmonica', 'backpack',
'steam_locomotive', 'ashcan', 'espresso_maker', 'buckle', 'water_jug', 'projector',
'seat_belt', 'motor_scooter', 'mailbox', 'tractor', 'moving_van', 'amphibian','half_track','garbage_truck','tank']
label = decode_predictions(out)
predicted_List = label[0]
setFlagForImageDetection = False
for tupleX in predicted_List:
if setFlagForImageDetection == True:
return True
else:
for classNames in carList:
if classNames in tupleX[1]:
setFlagForImageDetection = True
return True
if setFlagForImageDetection == False:
return False
def prepare_img_256(img_path):
img = load_img(img_path, target_size=(256, 256)) # this is a PIL image
x = img_to_array(img) # this is a Numpy array with shape (3, 256, 256)
x = x.reshape((1,) + x.shape)/255
return x
def car_damage_gate(img_256, model):
print ("Validating that damage exists...")
with graph.as_default():
pred = model.predict(img_256)
if pred[0][0] < 1:
return True # print "Validation complete - proceed to location and severity determination"
else:
return False
def location_assessment(img_128, model):
print ("Determining location of damage...")
with graph.as_default():
# pred = model.predict(img_128)
prediction = model.predict(img_128)
# print(prediction[0])
if prediction[0][0] <=.9:
# print("Front Part Of car")
return "Front Part Of car"
else:
return"Rear Part Of vehicle"
def severity_assessment(img_256, model):
print("Determining severity of damage...")
with graph.as_default():
pred_severe = model.predict(img_256)
pred_label = np.argmax(pred_severe, axis=1)
d = {0: 'Minor', 1: 'Moderate', 2: 'Severe'}
for key in d:
if pred_label[0] == key:
return "{} ".format(d[key])
def engine(img_path , userid):
# with graph.as_default():
# VGG16_model = load_model(str(modelpath) + '/vgg16.h5')
# print ("VGG16_model loaded")
img_224 = prepare_img_224(img_path)
g1 = car_categories_gate(img_224, VGG16_model)
# g1 = True
print(g1)
if g1 is False:
result = {'gate1': 'Car validation check: ',
'gate1_result': 0,
'gate1_message': {0: 'Are you sure this is a picture of your car? Please retry your submission.',
1: 'Hint: Try zooming in/out, using a different angle or different lighting'},
'gate2': None,
'gate2_result': None,
'gate2_message': {0: None, 1: None},
'location': None,
'severity': None,
'final': 'Damage assessment unsuccessful!'}
return result
# #Car Make model and number
print('***************start****************************')
headers = {"Content-type": "application/json","X-Access-Token": "rI0eHteykKhi6Rh9L46ujqELNCsh6wK4WOPy"}
conn = httplib.HTTPSConnection("dev.sighthoundapi.com",context=ssl.SSLContext(ssl.PROTOCOL_TLSv1))
# To use a hosted image uncomment the following line and update the URL
#image_data = "http://www.blogcdn.com/www.autoblog.com/media/2011/08/lead2-2012-toyota-camry-se-fd.jpg"
# To use a local file uncomment the following line and update the path
image_data = base64.b64encode(open(img_path, "rb").read()).decode()
params = json.dumps({"image": image_data})
print('********')
conn.request("POST", "/v1/recognition?objectType=vehicle,licenseplate", params, headers)
print('****hagshj****')
response = conn.getresponse()
result = response.read()
vehicleJsonOut = json.loads(result)
print('***************stop****************************')
makeOfvehicle = ''
modelOfVehicle = ''
typeOfVehicle = ''
numberOfvehicle = ''
regionofVehicle = ''
colorOfVehicle = ''
numberPlateFound = 0
resultfoundFlag = True
try:
vehicleJsonOut['objects'][0]['vehicleAnnotation']
except:
resultfoundFlag = False
if resultfoundFlag == True:
if 'attributes' in vehicleJsonOut['objects'][0]['vehicleAnnotation']:
makeOfvehicle = vehicleJsonOut['objects'][0]['vehicleAnnotation']['attributes']['system']['make']['name']
modelOfVehicle = vehicleJsonOut['objects'][0]['vehicleAnnotation']['attributes']['system']['model']['name']
typeOfVehicle = vehicleJsonOut['objects'][0]['vehicleAnnotation']['attributes']['system']['vehicleType']
colorOfVehicle = vehicleJsonOut['objects'][0]['vehicleAnnotation']['attributes']['system']['color']['name']
print("Make of vehicle: " + makeOfvehicle )
print('Model of vehicle: ' + modelOfVehicle)
print('Vehiicle type : ' + typeOfVehicle)
print('vehicle color : ' + colorOfVehicle )
if 'licenseplate' in vehicleJsonOut['objects'][0]['vehicleAnnotation']:
numberOfvehicle = vehicleJsonOut['objects'][0]['vehicleAnnotation']['licenseplate']['attributes']['system']['string']['name']
regionofVehicle = vehicleJsonOut['objects'][0]['vehicleAnnotation']['licenseplate']['attributes']['system']['region']['name']
print('Car Number : ' + numberOfvehicle)
print('region : ' + regionofVehicle )
# makeOfvehicle = 'Audi'
# modelOfVehicle = 'Q7'
# typeOfVehicle = 'SUV'
# numberOfvehicle = ''
#regionofVehicle = ''
#colorOfVehicle = 'Red'
#numberPlateFound = 0
img_128 = prepare_img_128(img_path)
# with graph.as_default():
# DamgeDetection_model = load_model(str(modelpath) + '/car_damage-model_retrained.h5')
# print ("DamgeDetection_model loaded")
g2 = car_damage_gate(img_128, DamgeDetection_model)
print(g2)
if g2 is False:
result = {'gate1': 'Car validation check: ',
'gate1_result': 1,
'gate1_message': {0: None, 1: None},
'gate2': 'Damage presence check: ',
'gate2_result': 0,
'makeModelGate' : numberPlateFound ,
'gate2_message': {0: 'Are you sure that your car is damaged? Please retry your submission.'},
'location': None,
'severity': None,
'make': makeOfvehicle ,
'model' : modelOfVehicle ,
'color' : colorOfVehicle ,
'typeOfvehicle' : typeOfVehicle,
'numberplate' : numberOfvehicle ,
'region' : regionofVehicle ,
'final': 'Damage assessment unsuccessful!'}
return result
# img_128 = prepare_img_128(img_path)
# with graph.as_default():
# location_model = load_model(str(modelpath) + '/carLocationWithfrontAndBack3400Images.h5')
# print ("Location model loaded")
# severity_model = load_model(str(modelpath)+ '/car_damage_severity.h5')
# print ("Severity model loaded")
g3 = location_assessment(img_128, location_model)
print("Location of damage : " + g3)
img_256 = prepare_img_256(img_path)
g4 = severity_assessment(img_256, severity_model)
print ('Severity of damage : ' + g4)
print(userid)
customer_ID = 1001
print(customer_ID)
import psycopg2
conn = psycopg2.connect(database="TouchNpaid", user = "postgres", password = "Admin123", host = "10.109.32.131", port = "5454")
cur = conn.cursor()
query = " SELECT YOM , PRICE from POLICYINFO WHERE USERID = '{c}'".format(c = customer_ID)
cur.execute(query)
rows = cur.fetchall()
# print ("USERID = ", rows[0][0])
# print ("POLICYID = ", rows[0][1])
# print ("VEHICLENUM = ", rows[0][2])
# print ("MAKE = ", rows[0][3])
# print ("MODEL = ", rows[0][4] )
yearOfMan = int(rows[0][0])
print("Manufacturing Year: " + str(yearOfMan))
costOfCar = int(rows[0][1] )
print("Cost of car : $" + str(costOfCar))
conn.close()
depreCost = 5
currentYr = 2019
deductible = 250
segmentPercent = 0.125
percentDepre = (currentYr - yearOfMan) * depreCost
print('percentDepre : ' + str(percentDepre))
finalAfterDepre = ((100 - percentDepre) / 100 ) * costOfCar
print('finalAfter depreciation : ' + str(finalAfterDepre))
estimatedCost = 0
severitycheck = str(g4)
if 'Severe' in severitycheck:
print('sever loop')
estimatedCost = finalAfterDepre*segmentPercent
estimatedCost = estimatedCost - deductible
elif 'Moderate' in severitycheck :
print('moderate loop')
estimatedCost = finalAfterDepre*segmentPercent*0.66
estimatedCost = estimatedCost - deductible
else:
estimatedCost = finalAfterDepre*segmentPercent*0.33
estimatedCost = estimatedCost - deductible
print(estimatedCost)
result = {'gate1': 'Car validation check: ',
'gate1_result': 1,
'gate1_message': {0: None, 1: None},
'gate2': 'Damage presence check: ',
'gate2_result': 1,
'gate2_message': {0: None, 1: None},
'location': g3,
'severity': g4,
'modelGate' : numberPlateFound ,
'make': makeOfvehicle ,
'model' : modelOfVehicle ,
'color' : colorOfVehicle ,
'typeOfvehicle' : typeOfVehicle,
'numberplate' : numberOfvehicle ,
'region' : regionofVehicle ,
'estimatedcost' : '$ ' + str(estimatedCost) ,
'final': 'Damage assessment complete!'}
return result