检查输入时出错：预期conv2d_1_input具有形状（64、64、3），但数组具有形状（64、64、4）

Question

My model for classifying images is shown below.When using the predict function it is giving the below error:

ValueError: Error when checking input: expected conv2d_1_input to have shape (64, 64, 3) but got array with shape (64, 64, 4)

The model is as shown below:

# Importing the Keras libraries and packages
from keras.models import Sequential
from keras.layers import Conv2D
from keras.layers import MaxPooling2D
from keras.layers import Flatten
from keras.layers import Dense
# Initialising the CNN
classifier = Sequential()
# Step 1 - Convolution
classifier.add(Conv2D(32, (3, 3), input_shape = (64, 64, 3), activation = 'relu'))
# Step 2 - Pooling
classifier.add(MaxPooling2D(pool_size = (2, 2)))
# Adding a second convolutional layer
classifier.add(Conv2D(32, (3, 3), activation = 'relu'))
classifier.add(MaxPooling2D(pool_size = (2, 2)))
# Step 3 - Flattening
classifier.add(Flatten())
# Step 4 - Full connection
classifier.add(Dense(units = 128, activation = 'relu'))
classifier.add(Dense(units = 1, activation = 'sigmoid'))
# Compiling the CNN
classifier.compile(optimizer = 'adam', loss = 'binary_crossentropy', 
metrics = ['accuracy'])

# Part 2 - Fitting the CNN to the images
from keras.preprocessing.image import ImageDataGenerator
import pickle
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('Dataset/training_data',
target_size = (64, 64),batch_size = 32,class_mode = 'binary')
test_set = test_datagen.flow_from_directory('Dataset/test_data',
target_size = (64, 64),batch_size = 32,class_mode = 'binary')
classifier.fit_generator(training_set,
steps_per_epoch = 350,epochs = 2,validation_data = test_set,validation_steps 
= 101)

The predict function is shown below: I have used requests package since I want to use a url of an image for prediction.

import requests
from io import BytesIO
from PIL import Image
import numpy as np
from keras.preprocessing import image
import cv2

 url='http://answers.opencv.org/upfiles/logo_2.png'
 response = requests.get(url)
 img = Image.open(BytesIO(response.content))
 #file = cv2.imread(img)
 img = img.resize((64,64))
 x = image.img_to_array(img)
 x = np.expand_dims(x, axis=0)
 result = classifier.predict(x)
 #training_set.class_indices
 if result[0][0] == 1:
      prediction = 'signature'
 else:
      prediction = 'nonsignature'



 print(prediction)

Is there an alternate way using only one package instead of PIL and keras

Thanks for help!!

Answer 1

The error message speaks for itself. Your network expects images with 3 color channels ( RGB ), but you are providing it images with 4 channels. You are working with a png image, so the image is probably in RGBA format and the fourth channel is a transparancy channel.

PIL images can be converted to RGB format as follows:

img = img.convert(mode='RGB')
x = image.img_to_array(img)

x.shape
> (64, 64, 3)

However, this conversion might not give the desired result (the background will be black for instance). You can refer to this question for other ways of converting to RGB .

You can also use the load_img function from the Keras preprocessing module:

import keras
img = keras.preprocessing.image.load_img(io.BytesIO(response.content))

This will load the image in RGB format. This function uses PIL under the hood and produces the same result as above.