As we mentioned earlier, we will be using the pre-trained inception model on the ImageNet dataset. So, we need to download this pre-trained model from the internet.
Let's start off by defining data_dir for the inception model:
inception.data_dir = 'inception/'The weights of the pre-trained inception model are about 85 MB. The following line of code will download it if it doesn't exist in the data_dir defined previously:
inception.maybe_download()
Downloading Inception v3 Model ...
- Download progress: 100%
We will load the inception model so that we can use it as a feature extractor for our CIFAR-10 images:
# Loading the inception model so that we can inialized it with the pre-trained weights and customize for our model
inception_model = inception.Inception()As we mentioned previously, calculating the transfer values for the CIFAR-10 dataset will take some time, so we need to cache them for future use. Thankfully, there's a helper function in the inception module that can help us do that:
from inception import transfer_values_cache
Next up, we need to set the file paths for the cached training and testing files:
file_path_train = os.path.join(cifar10.data_path, 'inception_cifar10_train.pkl')
file_path_test = os.path.join(cifar10.data_path, 'inception_cifar10_test.pkl')
print("Processing Inception transfer-values for the training images of Cifar-10 ...")
# First we need to scale the imgs to fit the Inception model requirements as it requires all pixels to be from 0 to 255,
# while our training examples of the CIFAR-10 pixels are between 0.0 and 1.0
imgs_scaled = training_images * 255.0
# Checking if the transfer-values for our training images are already calculated and loading them, if not calculate and save them.
transfer_values_training = transfer_values_cache(cache_path=file_path_train,
images=imgs_scaled,
model=inception_model)
print("Processing Inception transfer-values for the testing images of Cifar-10 ...")
# First we need to scale the imgs to fit the Inception model requirements as it requires all pixels to be from 0 to 255,
# while our training examples of the CIFAR-10 pixels are between 0.0 and 1.0
imgs_scaled = testing_images * 255.0
# Checking if the transfer-values for our training images are already calculated and loading them, if not calcaulate and save them.
transfer_values_testing = transfer_values_cache(cache_path=file_path_test,
images=imgs_scaled,
model=inception_model)
As mentioned before, we have 50,000 images in the training set of the CIFAR-10 dataset. So let's check the shapes of the transfer values of these images. It should be 2,048 for each image in this training set:
transfer_values_training.shape
Output:
(50000, 2048)
We need to do the same for the test set:
transfer_values_testing.shape
Output:
To intuitively understand how the transfer values look, we are going to define a helper function to enable us to use the plot the transfer values of a specific image from the training or the testing sets:
def plot_transferValues(ind):
print("Original input image:")
# Plot the image at index ind of the test set.
plt.imshow(testing_images[ind], interpolation='nearest')
plt.show()
print("Transfer values using Inception model:")
# Visualize the transfer values as an image.
transferValues_img = transfer_values_testing[ind]
transferValues_img = transferValues_img.reshape((32, 64))
# Plotting the transfer values image.
plt.imshow(transferValues_img, interpolation='nearest', cmap='Reds')
plt.show()
plot_transferValues(i=16)
Input image:
Figure 10.5: Input image
Transfer values for the image using the inception model:
plot_transferValues(i=17)
Transfer values for the image using the inception model:
