!wget --no-check-certificate \
--2019-05-23 08:11:52-- https://storage.googleapis.com/laurencemoroney-blog.appspot.com/horse-or-human.zip Resolving storage.googleapis.com (storage.googleapis.com)... 220.127.116.11, 2404:6800:4008:c06::80 Connecting to storage.googleapis.com (storage.googleapis.com)|18.104.22.168|:443... connected. HTTP request sent, awaiting response... 200 OK Length: 149574867 (143M) [application/zip] Saving to: ‘/tmp/horse-or-human.zip’ /tmp/horse-or-human 100%[===================>] 142.65M 81.4MB/s in 1.8s 2019-05-23 08:11:54 (81.4 MB/s) - ‘/tmp/horse-or-human.zip’ saved [149574867/149574867]
!wget --no-check-certificate \
--2019-05-23 08:13:02-- https://storage.googleapis.com/laurencemoroney-blog.appspot.com/validation-horse-or-human.zip Resolving storage.googleapis.com (storage.googleapis.com)... 22.214.171.124, 2404:6800:4008:c06::80 Connecting to storage.googleapis.com (storage.googleapis.com)|126.96.36.199|:443... connected. HTTP request sent, awaiting response... 200 OK Length: 11480187 (11M) [application/zip] Saving to: ‘/tmp/validation-horse-or-human.zip’ /tmp/validation-hor 100%[===================>] 10.95M --.-KB/s in 0.1s 2019-05-23 08:13:03 (96.6 MB/s) - ‘/tmp/validation-horse-or-human.zip’ saved [11480187/11480187]
The following python code will use the OS library to use Operating System libraries, giving you access to the file system, and the zipfile library allowing you to unzip the data.
The contents of the .zip are extracted to the base directory
/tmp/horse-or-human, which in turn each contain
In short: The training set is the data that is used to tell the neural network model that ‘this is what a horse looks like’, ‘this is what a human looks like’ etc.
One thing to pay attention to in this sample: We do not explicitly label the images as horses or humans. If you remember with the handwriting example earlier, we had labelled ‘this is a 1’, ‘this is a 7’ etc. Later you’ll see something called an ImageGenerator being used — and this is coded to read images from subdirectories, and automatically label them from the name of that subdirectory. So, for example, you will have a ‘training’ directory containing a ‘horses’ directory and a ‘humans’ one. ImageGenerator will label the images appropriately for you, reducing a coding step.
Let’s define each of these directories:
# Directory with our training horse pictures
Now, let’s see what the filenames look like in the
humans training directories:
train_horse_names = os.listdir(train_horse_dir)
['horse29-9.png', 'horse15-9.png', 'horse05-0.png', 'horse30-0.png', 'horse35-4.png', 'horse02-3.png', 'horse31-8.png', 'horse09-5.png', 'horse03-9.png', 'horse37-2.png'] ['human13-16.png', 'human08-24.png', 'human14-00.png', 'human16-16.png', 'human06-27.png', 'human04-17.png', 'human03-18.png', 'human16-00.png', 'human16-17.png', 'human09-24.png'] ['horse1-276.png', 'horse4-530.png', 'horse2-040.png', 'horse2-183.png', 'horse2-201.png', 'horse1-554.png', 'horse6-218.png', 'horse3-011.png', 'horse6-004.png', 'horse3-326.png'] ['valhuman02-13.png', 'valhuman01-23.png', 'valhuman03-24.png', 'valhuman01-07.png', 'valhuman02-14.png', 'valhuman05-08.png', 'valhuman03-12.png', 'valhuman05-27.png', 'valhuman04-10.png', 'valhuman05-11.png']
Let’s find out the total number of horse and human images in the directories:
print('total training horse images:', len(os.listdir(train_horse_dir)))
total training horse images: 500 total training human images: 527 total validation horse images: 128 total validation human images: 128
Now let’s take a look at a few pictures to get a better sense of what they look like. First, configure the matplot parameters:
Now, display a batch of 8 horse and 8 human pictures. You can rerun the cell to see a fresh batch each time:
# Set up matplotlib fig, and size it to fit 4x4 pics
But before we continue, let’s start defining the model:
Step 1 will be to import tensorflow.
import tensorflow as tf
We then add convolutional layers as in the previous example, and flatten the final result to feed into the densely connected layers.
Finally we add the densely connected layers.
Note that because we are facing a two-class classification problem, i.e. a binary classification problem, we will end our network with a sigmoid activation, so that the output of our network will be a single scalar between 0 and 1, encoding the probability that the current image is class 1 (as opposed to class 0).
model = tf.keras.models.Sequential([
WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/resource_variable_ops.py:435: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version. Instructions for updating: Colocations handled automatically by placer.
The model.summary() method call prints a summary of the NN
_________________________________________________________________ Layer (type) Output Shape Param # ================================================================= conv2d (Conv2D) (None, 298, 298, 16) 448 _________________________________________________________________ max_pooling2d (MaxPooling2D) (None, 149, 149, 16) 0 _________________________________________________________________ conv2d_1 (Conv2D) (None, 147, 147, 32) 4640 _________________________________________________________________ max_pooling2d_1 (MaxPooling2 (None, 73, 73, 32) 0 _________________________________________________________________ conv2d_2 (Conv2D) (None, 71, 71, 64) 18496 _________________________________________________________________ max_pooling2d_2 (MaxPooling2 (None, 35, 35, 64) 0 _________________________________________________________________ conv2d_3 (Conv2D) (None, 33, 33, 64) 36928 _________________________________________________________________ max_pooling2d_3 (MaxPooling2 (None, 16, 16, 64) 0 _________________________________________________________________ conv2d_4 (Conv2D) (None, 14, 14, 64) 36928 _________________________________________________________________ max_pooling2d_4 (MaxPooling2 (None, 7, 7, 64) 0 _________________________________________________________________ flatten (Flatten) (None, 3136) 0 _________________________________________________________________ dense (Dense) (None, 512) 1606144 _________________________________________________________________ dense_1 (Dense) (None, 1) 513 ================================================================= Total params: 1,704,097 Trainable params: 1,704,097 Non-trainable params: 0 _________________________________________________________________
The “output shape” column shows how the size of your feature map evolves in each successive layer. The convolution layers reduce the size of the feature maps by a bit due to padding, and each pooling layer halves the dimensions.
Next, we’ll configure the specifications for model training. We will train our model with the
binary_crossentropy loss, because it’s a binary classification problem and our final activation is a sigmoid. (For a refresher on loss metrics, see the Machine Learning Crash Course.) We will use the
rmsprop optimizer with a learning rate of
0.001. During training, we will want to monitor classification accuracy.
NOTE: In this case, using the RMSprop optimization algorithm is preferable to stochastic gradient descent (SGD), because RMSprop automates learning-rate tuning for us. (Other optimizers, such as Adam and Adagrad, also automatically adapt the learning rate during training, and would work equally well here.)
from tensorflow.keras.optimizers import RMSprop
Let’s set up data generators that will read pictures in our source folders, convert them to
float32 tensors, and feed them (with their labels) to our network. We’ll have one generator for the training images and one for the validation images. Our generators will yield batches of images of size 300x300 and their labels (binary).
As you may already know, data that goes into neural networks should usually be normalized in some way to make it more amenable to processing by the network. (It is uncommon to feed raw pixels into a convnet.) In our case, we will preprocess our images by normalizing the pixel values to be in the
[0, 1] range (originally all values are in the
[0, 255] range).
In Keras this can be done via the
keras.preprocessing.image.ImageDataGenerator class using the
rescale parameter. This
ImageDataGenerator class allows you to instantiate generators of augmented image batches (and their labels) via
.flow(data, labels) or
.flow_from_directory(directory). These generators can then be used with the Keras model methods that accept data generators as inputs:
from tensorflow.keras.preprocessing.image import ImageDataGenerator
Found 1027 images belonging to 2 classes. Found 256 images belonging to 2 classes.
Let’s train for 15 epochs — this may take a few minutes to run.
Do note the values per epoch.
The Loss and Accuracy are a great indication of progress of training. It’s making a guess as to the classification of the training data, and then measuring it against the known label, calculating the result. Accuracy is the portion of correct guesses.
history = model.fit_generator(
WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version. Instructions for updating: Use tf.cast instead. Epoch 1/15 8/8 [==============================] - 2s 216ms/step - loss: 0.6814 - acc: 0.5000 9/9 [==============================] - 11s 1s/step - loss: 0.8579 - acc: 0.5268 - val_loss: 0.6814 - val_acc: 0.5000 Epoch 2/15 8/8 [==============================] - 2s 206ms/step - loss: 0.8800 - acc: 0.5000 9/9 [==============================] - 9s 955ms/step - loss: 0.6284 - acc: 0.5881 - val_loss: 0.8800 - val_acc: 0.5000 Epoch 3/15 8/8 [==============================] - 2s 207ms/step - loss: 0.6334 - acc: 0.5781 9/9 [==============================] - 8s 940ms/step - loss: 0.6655 - acc: 0.6602 - val_loss: 0.6334 - val_acc: 0.5781 Epoch 4/15 8/8 [==============================] - 2s 207ms/step - loss: 0.4119 - acc: 0.8711 9/9 [==============================] - 8s 938ms/step - loss: 0.5446 - acc: 0.7790 - val_loss: 0.4119 - val_acc: 0.8711 Epoch 5/15 8/8 [==============================] - 2s 206ms/step - loss: 1.3072 - acc: 0.8125 9/9 [==============================] - 9s 956ms/step - loss: 0.4118 - acc: 0.8189 - val_loss: 1.3072 - val_acc: 0.8125 Epoch 6/15 8/8 [==============================] - 2s 205ms/step - loss: 2.0815 - acc: 0.7852 9/9 [==============================] - 9s 960ms/step - loss: 0.1871 - acc: 0.9309 - val_loss: 2.0815 - val_acc: 0.7852 Epoch 7/15 8/8 [==============================] - 2s 208ms/step - loss: 1.7291 - acc: 0.7539 9/9 [==============================] - 9s 1s/step - loss: 0.6104 - acc: 0.8384 - val_loss: 1.7291 - val_acc: 0.7539 Epoch 8/15 8/8 [==============================] - 2s 211ms/step - loss: 1.3320 - acc: 0.8242 9/9 [==============================] - 9s 956ms/step - loss: 0.1349 - acc: 0.9406 - val_loss: 1.3320 - val_acc: 0.8242 Epoch 9/15 8/8 [==============================] - 2s 205ms/step - loss: 1.7191 - acc: 0.7969 9/9 [==============================] - 8s 942ms/step - loss: 0.1120 - acc: 0.9581 - val_loss: 1.7191 - val_acc: 0.7969 Epoch 10/15 8/8 [==============================] - 2s 201ms/step - loss: 0.3554 - acc: 0.9062 9/9 [==============================] - 8s 938ms/step - loss: 0.0666 - acc: 0.9718 - val_loss: 0.3554 - val_acc: 0.9062 Epoch 11/15 8/8 [==============================] - 2s 205ms/step - loss: 1.1811 - acc: 0.8398 9/9 [==============================] - 8s 941ms/step - loss: 0.1556 - acc: 0.9669 - val_loss: 1.1811 - val_acc: 0.8398 Epoch 12/15 8/8 [==============================] - 2s 204ms/step - loss: 1.2945 - acc: 0.8398 9/9 [==============================] - 8s 938ms/step - loss: 0.0237 - acc: 0.9932 - val_loss: 1.2945 - val_acc: 0.8398 Epoch 13/15 8/8 [==============================] - 2s 203ms/step - loss: 1.0270 - acc: 0.8945 9/9 [==============================] - 9s 969ms/step - loss: 0.0034 - acc: 1.0000 - val_loss: 1.0270 - val_acc: 0.8945 Epoch 14/15 8/8 [==============================] - 2s 203ms/step - loss: 0.8244 - acc: 0.7773 9/9 [==============================] - 8s 940ms/step - loss: 0.5428 - acc: 0.9007 - val_loss: 0.8244 - val_acc: 0.7773 Epoch 15/15 8/8 [==============================] - 2s 207ms/step - loss: 1.0451 - acc: 0.8633 9/9 [==============================] - 9s 946ms/step - loss: 0.0516 - acc: 0.9873 - val_loss: 1.0451 - val_acc: 0.8633
Let’s now take a look at actually running a prediction using the model. This code will allow you to choose 1 or more files from your file system, it will then upload them, and run them through the model, giving an indication of whether the object is a horse or a human.
import numpy as np
Saving zhenjianzhao.jpg to zhenjianzhao (3).jpg [0.] zhenjianzhao.jpg is a horse
To get a feel for what kind of features our convnet has learned, one fun thing to do is to visualize how an input gets transformed as it goes through the convnet.
Let’s pick a random image from the training set, and then generate a figure where each row is the output of a layer, and each image in the row is a specific filter in that output feature map. Rerun this cell to generate intermediate representations for a variety of training images.
import numpy as np
/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:43: RuntimeWarning: invalid value encountered in true_divide
As you can see we go from the raw pixels of the images to increasingly abstract and compact representations. The representations downstream start highlighting what the network pays attention to, and they show fewer and fewer features being “activated”; most are set to zero. This is called “sparsity.” Representation sparsity is a key feature of deep learning.
These representations carry increasingly less information about the original pixels of the image, but increasingly refined information about the class of the image. You can think of a convnet (or a deep network in general) as an information distillation pipeline.
Before running the next exercise, run the following cell to terminate the kernel and free memory resources:
import os, signal