CNN模型只预测头等舱

Question

我正在研究对汽车模型进行细粒度分类的方法。所以我使用了迁移学习ResNet50。据我所知，它在训练时表现良好。但当我尝试新图像时，它总是预测一个类别。下面是我的代码。

培训要求:

from tensorflow.keras.layers import Input, Lambda, Dense, Flatten
from tensorflow.keras.models import Model
from tensorflow.keras.applications.resnet50 import ResNet50
from keras.applications.vgg16 import VGG16
from tensorflow.keras.applications.resnet50 import preprocess_input
from tensorflow.keras.preprocessing import image
from tensorflow.keras.preprocessing.image import ImageDataGenerator,load_img
from tensorflow.keras.models import Sequential
import matplotlib.pyplot as plt
import numpy as np
from glob import glob

IMAGE_SIZE = [224, 224]
train_path = 'Datasets/train'
valid_path = 'Datasets/test'
resnet = ResNet50(input_shape = IMAGE_SIZE + [3], weights='imagenet', include_top = False)

for layer in resnet.layers:
layer.trainable = False
folders = glob('Datasets/train/*') #training folders
x = Flatten()(resnet.output)
prediction = Dense(len(folders), activation='softmax') (x) 
model = Model(inputs = resnet.input, outputs = prediction)
model.compile(
loss = 'categorical_crossentropy',
optimizer = 'adam',
metrics = ['accuracy']
)
train_datagen = ImageDataGenerator(rescale = 1./255,
shear_range = 0.2,
zoom_range = 0.2,
horizontal_flip = True)
test_datagen = ImageDataGenerator(rescale = 1./255)
training_set = train_datagen.flow_from_directory('Datasets/train',
target_size = (224, 224),
batch_size = 32,
class_mode = 'categorical')

test_set = test_datagen.flow_from_directory('Datasets/test',
target_size = (224, 224),
batch_size = 32,
class_mode = 'categorical')

r = model.fit_generator(
training_set,
validation_data=test_set,
epochs=200,
steps_per_epoch=len(training_set),
validation_steps=len(test_set)
)

from tensorflow.keras.models import load_model
model.save('model_updateV1.h5')
y_pred = model.predict(test_set)

import numpy as np
y_pred = np.argmax(y_pred, axis=1)

用于尝试新图像:

from tensorflow.keras.models import load_model
from tensorflow.keras.preprocessing import image
import numpy as np
from tensorflow.keras.applications.resnet50 import preprocess_input
model = load_model('model_updateV1.h5')
img = image.load_img('Datasets/test/mercedes/45.jpg', target_size=(224,224))
x = image.img_to_array(img)
x = x/255.
x = np.expand_dims(x, axis = 0)
img_data = preprocess_input(x)
img_data.shape
model.predict(img_data)
a = np.argmax(model.predict(img_data), axis=1)
a

Answer 1

我认为你的问题是你重新缩放图像两次。你有代码

x=x/255

则展开维度，这很好。然后你有代码

img_data = preprocess_input(x)

我认为preprocess_input函数用代码 重新缩放-1和+1之间的像素值

x=x/127.5-1. 

现在你的像素值已经缩小了两倍。因此只需删除代码

x=x/255