def Example():
# image_path = args.data_path + '\\20190301074833_Wave1_2019Y03M01D_07h48m33s_3sec_nonhighpass.png'
image_path = 'D:\\Onepredict_MK\\LG CNS\\20190417_LG_CNS_test_mode\\puppy.jpg'
resnet = ResNet50(input_shape=(224, 224, 3), weights='imagenet', include_top=True)
print(resnet.summary())
activation_layer = resnet.get_layer('activation_48')
model = Model(inputs=resnet.input, outputs=activation_layer.output)
final_dense = resnet.get_layer('fc1000')
weight = final_dense.get_weights()[0]
img = image.load_img(image_path, target_size=(224, 224))
img_input = preprocess_input(np.expand_dims(img, 0))
fmaps = model.predict(img_input)[0] # fmaps.shape = (7, 7, 2048)
probs = resnet.predict(img_input) # probs.shape = (1, 1000)
class_names = decode_predictions(probs)
class_name = class_names[0][0]
pred = np.argmax(probs[0]) # pred = 207
w = weight[:, pred] # w.shape = (2048,)
cam = fmaps.dot(w) # cam.shape = (7, 7)
camp = ndimage.zoom(cam, (32, 32), order=1)
plt.subplot(1, 2, 1)
plt.imshow(img, alpha=0.8)
plt.imshow(camp, cmap='jet', alpha=0.5)
plt.subplot(1, 2, 2)
plt.imshow(img)
plt.title(class_name)
plt.show()
def predict(img_local_path):
model = SqueezeNet(weights='imagenet')
img = image.load_img(img_local_path, target_size=(227, 227))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
preds = model.predict(x)
result = decode_predictions(preds)
return result
def train(img_path):
K.clear_session()
my_model = ResNet50(
weights='./resnet50_weights_tf_dim_ordering_tf_kernels.h5')
image_size = 224
img = load_img(img_path, target_size=(image_size, image_size))
img_array = np.array([img_to_array(img)])
inpu = preprocess_input(img_array)
preds = my_model.predict(inpu)
deco = decode_predictions(preds)
return deco
"""
@author: 代码医生工作室
@公众号：xiangyuejiqiren （内有更多优秀文章及学习资料）
@来源: <深度学习之TensorFlow工程化项目实战>配套代码 （700+页）
@配套代码技术支持：bbs.aianaconda.com (有问必答)
"""
from tensorflow.python.keras.applications.resnet50 import ResNet50
from tensorflow.python.keras.preprocessing import image
from tensorflow.python.keras.applications.resnet50 import preprocess_input, decode_predictions
import numpy as np
#6.7.10节 动态图
#tf.enable_eager_execution() #启动动态图
model = ResNet50(weights='imagenet')
#model = ResNet50(weights='resnet50_weights_tf_dim_ordering_tf_kernels.h5')
img_path = 'hy.jpg'
img = image.load_img(img_path, target_size=(224, 224))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
preds = model.predict(x)
print('Predicted:', decode_predictions(preds, top=3)[0])
#6.7.10节 动态图
#preds = model(x)
#print('Predicted:', decode_predictions(preds.numpy(), top=3)[0])
with tf.Session() as sess: #在session中运行
sess.run(tf.global_variables_initializer())
Reslayer = ResNet50(
weights='resnet50_weights_tf_dim_ordering_tf_kernels.h5')
logits = Reslayer(tensorimg) #网络模型
#Reslayer = ResNet50(weights='resnet50_weights_tf_dim_ordering_tf_kernels.h5',
# input_tensor=tensorimg,input_shape = (224, 224, 3) )
#logits = Reslayer.layers[-1].output
#print(logits)
img_path = './dog.jpg'
img = image.load_img(img_path, target_size=(224, 224))
logitsv = sess.run(logits, feed_dict={inputs: img})
Pred = decode_predictions(logitsv, top=3)[0]
print('Predicted:', Pred, len(logitsv[0]))
#可视化
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(10, 8))
fig.sca(ax1)
ax1.imshow(img)
fig.sca(ax1)
barlist = ax2.bar(range(3), [i[2] for i in Pred])
barlist[0].set_color('g')
plt.sca(ax2)
plt.ylim([0, 1.1])
#plt.xticks(range(3),[i[1][:15] for i in Pred], rotation='vertical')
fig.subplots_adjust(bottom=0.2)
def predictNevaluate(model, target_size, cwd, categ_names, wnids_list):
'''
target_size is a tuple of image-dimensions expected from every model
wnids_list is nested list with url's ending for each category, and the corresponding categ name.
wnids_list and categ_names must have the same sequnce, and matters to confusion matrix computation.
categ_names is list of categories. Their sequnce matters to confusion matrix computation.
Returns a dataframe
'''
# Create Confusion Matrix
# Create dataframe full of zeros to store confusion matrix
cm = pd.DataFrame(np.zeros((len(wnids_list)+1, len(wnids_list)+2)))
# Create column-names for confusion matrix
cm.columns = categ_names+['Other', 'Precision-Producer-Accur'] # Sorted categories
# Create row-names for confusion matrix
cm.index = categ_names+['Recall-User-Reliab']
for i in range(len(wnids_list)):
img_path = cwd +'/'+ wnids_list[i][0]
print('\n\nImages from folder {}:'.format(wnids_list[i][0]))
for root, subfolders, files in os.walk(img_path):
for file in files:
# load an image in PIL format
img = image.load_img(img_path+'/'+file, target_size=target_size)
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
preds = model.predict(x)
# Top three predictions for every image
imRes = decode_predictions(preds, top=3)[0]
# decode the results into a list of tuples (class, description, probability)
print('\nFile {} predicted:\n{}'.format(file, imRes))
# Create confusion matrix
# If name of predicted category is same as folder. Compare strings
if imRes[0][1] == wnids_list[i][0]:
# Add 1 at value in cell with row same as folder name
whichRow = np.where(cm.index == wnids_list[i][0])
# and col as predicted-name. [col][row]
cm[imRes[0][1]][whichRow[0][0]] += 1
print('TRUE\n')
# If is different from folder but is one of the five categories
elif (imRes[0][1] in categ_names and imRes[0][1] != wnids_list[i][0]):
# Add 1 at value in cell with row same as folder name
whichRow = np.where(cm.index == wnids_list[i][0])
# and col as predicted-name.
cm[imRes[0][1]][whichRow[0][0]] += 1
print('FALSE\n')
# If is none of the five categories
elif imRes[0][1] not in categ_names:
# Add 1 at value in cell with row same as folder name
whichRow = np.where(cm.index == wnids_list[i][0])
# and column as classified name, here 'Others'.
cm['Other'][whichRow[0][0]] += 1
print('OTHER\n')
# Compute Acc and Rel in Confusion Matrix
# Rel
for categ in cm.columns[:-1]: # For every (column) category (and 'other')
cm[categ][-1] = sum(cm[categ][:-1]) # Compute Total sum of column and write to last row [col][row]
if categ != 'Other': # Only for real categories, compute Rel and replace value in last row
# rel = diagonal element / last row element
rel = (cm[categ][categ_names.index(categ)] / cm[categ][-1])
# assign rel-value to last row
cm[categ][-1] = round(rel, 2)
# Acc
for categ in cm.index[:-1]: # For every (row) category (Here we dont have 'other')
whichRow = np.where(cm.index == categ) # row is where index == categ name
rowValues = cm.iloc[whichRow[0][0]][:-1] # select the whole row
# Acc = diagonal element / sum of row
acc = cm[categ][categ_names.index(categ)] / sum(rowValues)
# assign acc-value to last column
cm[cm.columns[-1]][categ_names.index(categ)] = round(acc, 2)
# f1-score
for categ in cm.index[:-1]: # For every (row) category (Here we dont have 'other')
# Rel ^-1 (inverse)
whichRow = np.where(cm.index == 'Recall-User-Reliab') # row is where index == Rel
relValues = cm.iloc[whichRow[0][0]][:-2] # select the whole row - rowValues
invRelValues = [1/x for x in relValues]
# Acc ^-1 (inverse)
accValues = cm['Precision-Producer-Accur'][:-1]
invAccValues = [1/x for x in accValues]
# f1-score
f1 = (len(invRelValues)+len(invAccValues)) / (sum(invRelValues)+sum(invAccValues))
# Assign f1 to bottom right element
cm.iloc[whichRow[0][0]][-1] = round(f1, 2)
return cm
from tensorflow.python.keras.preprocessing import image
from tensorflow.python.keras.applications.resnet50 import ResNet50
from tensorflow.python.keras.applications.resnet50 import preprocess_input, decode_predictions
# Step 1: Preprocess data
img = image.load_img("./img/41cqodicvbl.jpg", target_size=(224, 224))
# Convert the image to a numpy array
img = image.img_to_array(img)
# Add a forth dimension since Keras expects a list of images
img = np.expand_dims(img, axis=0)
# Step 2: Load Pre-trained Model
model = ResNet50()
# Scale the input image to the range used in the trained network
img = preprocess_input(img)
# Run the image through the deep neural network to make a prediction
predictions = model.predict(img)
# Look up the names of the predicted classes. Index zero is the results for the first image.
predicted_classes = decode_predictions(predictions, top=3)
print("This is an image of:")
for imagenet_id, name, likelihood in predicted_classes[0]:
print(" - {}: {:2f} likelihood".format(name, likelihood * 100))
#-*-encoding:utf-8-*-
import numpy as np
from tensorflow.python.keras.preprocessing import image
from tensorflow.python.keras.applications import resnet50
import warnings
warnings.filterwarnings("ignore")
img = image.load_img('dog.png')
# print(image.img_to_array(img).shape)
model = resnet50.ResNet50(weights='imagenet')
img = image.load_img('dog.png', target_size=(224, 224))
img = image.img_to_array(img)
img = np.expand_dims(img, axis=0)
print(img.shape)
img = resnet50.preprocess_input(img)
#模型预测
pred = model.predict(img)
n = 10
top_n = resnet50.decode_predictions(pred, n)
for i in top_n[0]:
print(i)
# this classifier can be used with a bunch of different models from tensorflow.python.keras.applications.resnet50 import ResNet50 from tensorflow.python.keras.preprocessing import image from tensorflow.python.keras.applications.resnet50 import preprocess_input, decode_predictions import numpy as np # ResNet50 keras model can classify animals, model = ResNet50() img_path = '/home/vince/Desktop/honeybadger.jpg' img = image.load_img(img_path, target_size=(224, 224)) x = image.img_to_array(img) x = np.expand_dims(x, axis=0) x = preprocess_input(x) preds = model.predict(x) print(decode_predictions(preds, top=4))