# Artificial Intelligence Nanodegree¶

## Project: Write an Algorithm for a Dog Identification App¶

In this notebook, some template code has already been provided for you, and you will need to implement additional functionality to successfully complete this project. You will not need to modify the included code beyond what is requested. Sections that begin with '(IMPLEMENTATION)' in the header indicate that the following block of code will require additional functionality which you must provide. Instructions will be provided for each section, and the specifics of the implementation are marked in the code block with a 'TODO' statement. Please be sure to read the instructions carefully!

Note: Once you have completed all of the code implementations, you need to finalize your work by exporting the iPython Notebook as an HTML document. Before exporting the notebook to html, all of the code cells need to have been run so that reviewers can see the final implementation and output. You can then export the notebook by using the menu above and navigating to \n", "File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.

Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. Markdown cells can be edited by double-clicking the cell to enter edit mode.

The rubric contains optional "Stand Out Suggestions" for enhancing the project beyond the minimum requirements. If you decide to pursue the "Stand Out Suggestions", you should include the code in this IPython notebook.

### Why We're Here¶

In this notebook, you will make the first steps towards developing an algorithm that could be used as part of a mobile or web app. At the end of this project, your code will accept any user-supplied image as input. If a dog is detected in the image, it will provide an estimate of the dog's breed. If a human is detected, it will provide an estimate of the dog breed that is most resembling. The image below displays potential sample output of your finished project (... but we expect that each student's algorithm will behave differently!).

In this real-world setting, you will need to piece together a series of models to perform different tasks; for instance, the algorithm that detects humans in an image will be different from the CNN that infers dog breed. There are many points of possible failure, and no perfect algorithm exists. Your imperfect solution will nonetheless create a fun user experience!

We break the notebook into separate steps. Feel free to use the links below to navigate the notebook.

• Step 0: Import Datasets
• Step 1: Detect Humans
• Step 2: Detect Dogs
• Step 3: Create a CNN to Classify Dog Breeds (from Scratch)
• Step 4: Use a CNN to Classify Dog Breeds (using Transfer Learning)
• Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)
• Step 6: Write your Algorithm
• Step 7: Test Your Algorithm

## Step 0: Import Datasets¶

### Import Dog Dataset¶

In the code cell below, we import a dataset of dog images. We populate a few variables through the use of the load_files function from the scikit-learn library:

• train_files, valid_files, test_files - numpy arrays containing file paths to images
• train_targets, valid_targets, test_targets - numpy arrays containing onehot-encoded classification labels
• dog_names - list of string-valued dog breed names for translating labels
In [1]:
from sklearn.datasets import load_files
from keras.utils import np_utils
import numpy as np
from glob import glob

# define function to load train, test, and validation datasets
dog_files = np.array(data['filenames'])
dog_targets = np_utils.to_categorical(np.array(data['target']), 133)
return dog_files, dog_targets

# load train, test, and validation datasets

# load list of dog names
dog_names = [item[20:-1] for item in sorted(glob("dogImages/train/*/"))]

# print statistics about the dataset
print('There are %d total dog categories.' % len(dog_names))
print('There are %s total dog images.\n' % len(np.hstack([train_files, valid_files, test_files])))
print('There are %d training dog images.' % len(train_files))
print('There are %d validation dog images.' % len(valid_files))
print('There are %d test dog images.'% len(test_files))

Using TensorFlow backend.

There are 133 total dog categories.
There are 8351 total dog images.

There are 6680 training dog images.
There are 835 validation dog images.
There are 836 test dog images.


### Import Human Dataset¶

In the code cell below, we import a dataset of human images, where the file paths are stored in the numpy array human_files.

In [2]:
import random
random.seed(8675309)

# load filenames in shuffled human dataset
human_files = np.array(glob("lfw/*/*"))
random.shuffle(human_files)

# print statistics about the dataset
print('There are %d total human images.' % len(human_files))

There are 13233 total human images.


DEFINE Common Functions for use throughout

In [3]:
from sklearn.datasets import load_files
from keras.utils import np_utils
import numpy as np
from glob import glob
from keras.layers import Conv2D, MaxPooling2D, GlobalAveragePooling2D
from keras.layers import Dropout, Flatten, Dense
from keras.models import Sequential
from keras.callbacks import ModelCheckpoint  , EarlyStopping
from extract_bottleneck_features import *
import cv2
import matplotlib.pyplot as plt
%matplotlib inline

from keras.preprocessing import image
from tqdm import tqdm
import os

# VARIABLES

early_stop = EarlyStopping(patience=5)

def calculate_accuracy(f, true_files,false_files):
fp = 0
tp = 0
fn = 0
tn = 0
ttl = len(true_files) + len(false_files)

for h in true_files:
has_face = f(h)
if not has_face:
fn += 1
else: tp += 1

for d in false_files:
has_face = f(d)
if has_face:
fp += 1
else : tn += 1

recall = (tp / (tp + fp))*100
precision = (tp / (tp + fn ))*100
accuracy = (float(tp + tn) / ttl ) * 100

print("ACCURACY={}%\n\tRecall = {}%\n\tPrecision = {}%\n\tTrue Positives={}\n\tFalse Positives={}\n\tFalse Negatives={}\n\tTrue Negatives={}\n\tTOTAL={}".format( accuracy,
recall,
precision,
tp, fp, fn, tn, ttl))
def display_img(img_path):
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
plt.imshow(gray)
plt.show()

def path_to_tensor(img_path):
# loads RGB image as PIL.Image.Image type
# convert PIL.Image.Image type to 3D tensor with shape (224, 224, 3)
x = image.img_to_array(img)
# convert 3D tensor to 4D tensor with shape (1, 224, 224, 3) and return 4D tensor
return np.expand_dims(x, axis=0)

def paths_to_tensor(img_paths):
list_of_tensors = [path_to_tensor(img_path) for img_path in tqdm(img_paths)]
return np.vstack(list_of_tensors)

# define function to load train, test, and validation datasets
dog_files = np.array(data['filenames'])
dog_targets = np_utils.to_categorical(np.array(data['target']), 133)
return dog_files, dog_targets

from keras.wrappers.scikit_learn import KerasClassifier
from sklearn.model_selection import GridSearchCV

def create_model(optimizer,shape):
VGG19_model = Sequential()
VGG19_model.compile(loss='categorical_crossentropy', optimizer=optimizer, metrics=['accuracy'])
return VGG19_model


## Step 1: Detect Humans¶

We use OpenCV's implementation of Haar feature-based cascade classifiers to detect human faces in images. OpenCV provides many pre-trained face detectors, stored as XML files on github. We have downloaded one of these detectors and stored it in the haarcascades directory.

In the next code cell, we demonstrate how to use this detector to find human faces in a sample image.

In [17]:
import cv2
import matplotlib.pyplot as plt
%matplotlib inline

# extract pre-trained face detector

# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# find faces in image

# print number of faces detected in the image
print('Number of faces detected:', len(faces))

# get bounding box for each detected face
for (x,y,w,h) in faces:
# add bounding box to color image
cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)

# convert BGR image to RGB for plotting
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

# display the image, along with bounding box
plt.imshow(cv_rgb)
plt.show()

Number of faces detected: 3


Before using any of the face detectors, it is standard procedure to convert the images to grayscale. The detectMultiScale function executes the classifier stored in face_cascade and takes the grayscale image as a parameter.

In the above code, faces is a numpy array of detected faces, where each row corresponds to a detected face. Each detected face is a 1D array with four entries that specifies the bounding box of the detected face. The first two entries in the array (extracted in the above code as x and y) specify the horizontal and vertical positions of the top left corner of the bounding box. The last two entries in the array (extracted here as w and h) specify the width and height of the box.

### Write a Human Face Detector¶

We can use this procedure to write a function that returns True if a human face is detected in an image and False otherwise. This function, aptly named face_detector, takes a string-valued file path to an image as input and appears in the code block below.

In [4]:
# returns "True" if face is detected in image stored at img_path
def face_detector(img_path):
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
return len(faces) > 0


### (IMPLEMENTATION) Assess the Human Face Detector¶

Question 1: Use the code cell below to test the performance of the face_detector function.

• What percentage of the first 100 images in human_files have a detected human face?
• What percentage of the first 100 images in dog_files have a detected human face?

Ideally, we would like 100% of human images with a detected face and 0% of dog images with a detected face. You will see that our algorithm falls short of this goal, but still gives acceptable performance. We extract the file paths for the first 100 images from each of the datasets and store them in the numpy arrays human_files_short and dog_files_short.

In [5]:
human_files_short = human_files[:100]
dog_files_short = train_files[:100]
# Do NOT modify the code above this line.

## TODO: Test the performance of the face_detector algorithm
## on the images in human_files_short and dog_files_short.

# Here we define a function that will test all the faces using the first function face_detector on human_files and dog_files
# We calculate the accuracy of both the face detector on human files, then on dog files

calculate_accuracy(face_detector, human_files_short,dog_files_short)
calculate_accuracy(face_detector,dog_files_short, human_files_short)

ACCURACY=94.0%
Recall = 90.0%
Precision = 99.0%
True Positives=99
False Positives=11
False Negatives=1
True Negatives=89
TOTAL=200
ACCURACY=6.0%
Recall = 10.0%
Precision = 11.0%
True Positives=11
False Positives=99
False Negatives=89
True Negatives=1
TOTAL=200


Answer: As you can see the percentage for our Human files face detector is 94%, while the faces found on our dogs is only 6%.

Question 2: This algorithmic choice necessitates that we communicate to the user that we accept human images only when they provide a clear view of a face (otherwise, we risk having unneccessarily frustrated users!). In your opinion, is this a reasonable expectation to pose on the user? If not, can you think of a way to detect humans in images that does not necessitate an image with a clearly presented face?

I think it is a resaonable expectation and by clearly communicating that fact you can set up users expectations as well. There are off the shelf solutions that would not be difficult to implement so I see no problem using them.

Haar cascades give a great machine learning solution to calculating faces in images by using Adaboost , which uses a series of weaker classifiers trained on every feature, and takes a weighted sum of each individual classifier to come out with a final result. Boosting is a common and powerful technique to combine several predictive models and take the average ( often weighted ) of there outputs.

Another approach, if a face was not clearly presented, is to try various transformations on the original image , such as flipping and rotating, to see if that fixes the face not being easily detected. In the end, the quality of the picture will effect the prediction, which needs to be clearly communicated to the user.

## Step 2: Detect Dogs¶

In this section, we use a pre-trained ResNet-50 model to detect dogs in images. Our first line of code downloads the ResNet-50 model, along with weights that have been trained on ImageNet, a very large, very popular dataset used for image classification and other vision tasks. ImageNet contains over 10 million URLs, each linking to an image containing an object from one of 1000 categories. Given an image, this pre-trained ResNet-50 model returns a prediction (derived from the available categories in ImageNet) for the object that is contained in the image.

In [30]:
from keras.applications.resnet50 import ResNet50

# define ResNet50 model
ResNet50_model = ResNet50(weights='imagenet')


### Pre-process the Data¶

When using TensorFlow as backend, Keras CNNs require a 4D array (which we'll also refer to as a 4D tensor) as input, with shape

$$(\text{nb_samples}, \text{rows}, \text{columns}, \text{channels}),$$

where nb_samples corresponds to the total number of images (or samples), and rows, columns, and channels correspond to the number of rows, columns, and channels for each image, respectively.

The path_to_tensor function below takes a string-valued file path to a color image as input and returns a 4D tensor suitable for supplying to a Keras CNN. The function first loads the image and resizes it to a square image that is $224 \times 224$ pixels. Next, the image is converted to an array, which is then resized to a 4D tensor. In this case, since we are working with color images, each image has three channels. Likewise, since we are processing a single image (or sample), the returned tensor will always have shape

$$(1, 224, 224, 3).$$

The paths_to_tensor function takes a numpy array of string-valued image paths as input and returns a 4D tensor with shape

$$(\text{nb_samples}, 224, 224, 3).$$

Here, nb_samples is the number of samples, or number of images, in the supplied array of image paths. It is best to think of nb_samples as the number of 3D tensors (where each 3D tensor corresponds to a different image) in your dataset!

In [9]:
from keras.preprocessing import image
from tqdm import tqdm

# DEFINED ABOVE

# def path_to_tensor(img_path):
#     # loads RGB image as PIL.Image.Image type
#     img = image.load_img(img_path, target_size=(224, 224))
#     # convert PIL.Image.Image type to 3D tensor with shape (224, 224, 3)
#     x = image.img_to_array(img)
#     # convert 3D tensor to 4D tensor with shape (1, 224, 224, 3) and return 4D tensor
#     return np.expand_dims(x, axis=0)

# def paths_to_tensor(img_paths):
#     list_of_tensors = [path_to_tensor(img_path) for img_path in tqdm(img_paths)]
#     return np.vstack(list_of_tensors)


### Making Predictions with ResNet-50¶

Getting the 4D tensor ready for ResNet-50, and for any other pre-trained model in Keras, requires some additional processing. First, the RGB image is converted to BGR by reordering the channels. All pre-trained models have the additional normalization step that the mean pixel (expressed in RGB as $[103.939, 116.779, 123.68]$ and calculated from all pixels in all images in ImageNet) must be subtracted from every pixel in each image. This is implemented in the imported function preprocess_input. If you're curious, you can check the code for preprocess_input here.

Now that we have a way to format our image for supplying to ResNet-50, we are now ready to use the model to extract the predictions. This is accomplished with the predict method, which returns an array whose $i$-th entry is the model's predicted probability that the image belongs to the $i$-th ImageNet category. This is implemented in the ResNet50_predict_labels function below.

By taking the argmax of the predicted probability vector, we obtain an integer corresponding to the model's predicted object class, which we can identify with an object category through the use of this dictionary.

In [31]:
from keras.applications.resnet50 import preprocess_input, decode_predictions

def ResNet50_predict_labels(img_path):
# returns prediction vector for image located at img_path
img = preprocess_input(path_to_tensor(img_path))
return np.argmax(ResNet50_model.predict(img))


### Write a Dog Detector¶

While looking at the dictionary, you will notice that the categories corresponding to dogs appear in an uninterrupted sequence and correspond to dictionary keys 151-268, inclusive, to include all categories from 'Chihuahua' to 'Mexican hairless'. Thus, in order to check to see if an image is predicted to contain a dog by the pre-trained ResNet-50 model, we need only check if the ResNet50_predict_labels function above returns a value between 151 and 268 (inclusive).

We use these ideas to complete the dog_detector function below, which returns True if a dog is detected in an image (and False if not).

In [32]:
### returns "True" if a dog is detected in the image stored at img_path
def dog_detector(img_path):
prediction = ResNet50_predict_labels(img_path)
return ((prediction <= 268) & (prediction >= 151))


### (IMPLEMENTATION) Assess the Dog Detector¶

Question 3: Use the code cell below to test the performance of your dog_detector function.

• What percentage of the images in human_files_short have a detected dog?
• What percentage of the images in dog_files_short have a detected dog?

In [50]:
### TODO: Test the performance of the dog_detector function
### on the images in human_files_short and dog_files_short.

calculate_accuracy(dog_detector,dog_files_short,human_files_short)
calculate_accuracy(dog_detector,human_files_short,dog_files_short)

ACCURACY=100.0%
Recall = 100%
Precision = 100%
True Positives=100
False Positives=-1
False Negatives=-1
True Negatives=100
TOTAL=200
ACCURACY=0%
Recall = 0%
Precision = 0%
True Positives=-1
False Positives=100
False Negatives=100
True Negatives=0
TOTAL=200


Answer: Our dog_detector did much better than the face_detector, getting all 100% of the dogs in the dog files, and identified 0% of the humans as dogs

## Step 3: Create a CNN to Classify Dog Breeds (from Scratch)¶

Now that we have functions for detecting humans and dogs in images, we need a way to predict breed from images. In this step, you will create a CNN that classifies dog breeds. You must create your CNN from scratch (so, you can't use transfer learning yet!), and you must attain a test accuracy of at least 1%. In Step 5 of this notebook, you will have the opportunity to use transfer learning to create a CNN that attains greatly improved accuracy.

Be careful with adding too many trainable layers! More parameters means longer training, which means you are more likely to need a GPU to accelerate the training process. Thankfully, Keras provides a handy estimate of the time that each epoch is likely to take; you can extrapolate this estimate to figure out how long it will take for your algorithm to train.

We mention that the task of assigning breed to dogs from images is considered exceptionally challenging. To see why, consider that even a human would have great difficulty in distinguishing between a Brittany and a Welsh Springer Spaniel.

Brittany Welsh Springer Spaniel

It is not difficult to find other dog breed pairs with minimal inter-class variation (for instance, Curly-Coated Retrievers and American Water Spaniels).

Curly-Coated Retriever American Water Spaniel

Likewise, recall that labradors come in yellow, chocolate, and black. Your vision-based algorithm will have to conquer this high intra-class variation to determine how to classify all of these different shades as the same breed.

We also mention that random chance presents an exceptionally low bar: setting aside the fact that the classes are slightly imabalanced, a random guess will provide a correct answer roughly 1 in 133 times, which corresponds to an accuracy of less than 1%.

Remember that the practice is far ahead of the theory in deep learning. Experiment with many different architectures, and trust your intuition. And, of course, have fun!

### Pre-process the Data¶

We rescale the images by dividing every pixel in every image by 255.

In [13]:
from PIL import ImageFile

# pre-process the data for Keras
train_tensors = paths_to_tensor(train_files).astype('float32')/255
valid_tensors = paths_to_tensor(valid_files).astype('float32')/255
test_tensors = paths_to_tensor(test_files).astype('float32')/255

100%|██████████| 6680/6680 [00:53<00:00, 125.66it/s]
100%|██████████| 835/835 [00:05<00:00, 159.24it/s]
100%|██████████| 836/836 [00:05<00:00, 140.84it/s]


### (IMPLEMENTATION) Model Architecture¶

Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:

    model.summary()



We have imported some Python modules to get you started, but feel free to import as many modules as you need. If you end up getting stuck, here's a hint that specifies a model that trains relatively fast on CPU and attains >1% test accuracy in 5 epochs:

Question 4: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. If you chose to use the hinted architecture above, describe why you think that CNN architecture should work well for the image classification task.

In [14]:
from keras.models import Sequential
from keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, Dropout, GlobalAveragePooling2D

input_dims = (train_tensors.shape)[1:]
model = Sequential()

# The first three convolutional layers are to build up the features
# The Pooling layers are the key to reducing the size , the combination of the two layers increases depth

input_shape=input_dims))
# Bump to 32
# Bump to 64
# I added this in last minute as I had forgotten to originally.
# Adding this in increased accuracy , but seemed to also increase training times

#And finally our output layer, which will output probabilities for each one of the 133 breeds.

model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
conv2d_1 (Conv2D)            (None, 224, 224, 16)      208
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 112, 112, 16)      0
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 112, 112, 32)      2080
_________________________________________________________________
max_pooling2d_3 (MaxPooling2 (None, 56, 56, 32)        0
_________________________________________________________________
conv2d_3 (Conv2D)            (None, 56, 56, 65)        8385
_________________________________________________________________
max_pooling2d_4 (MaxPooling2 (None, 28, 28, 65)        0
_________________________________________________________________
global_average_pooling2d_1 ( (None, 65)                0
_________________________________________________________________
dense_1 (Dense)              (None, 133)               8778
=================================================================
Total params: 19,451
Trainable params: 19,451
Non-trainable params: 0
_________________________________________________________________


### Compile the Model¶

In [15]:
model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])


### (IMPLEMENTATION) Train the Model¶

Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.

You are welcome to augment the training data, but this is not a requirement.

In [16]:
### TODO: specify the number of epochs that you would like to use to train the model.

epochs = 20

### Do NOT modify the code below this line.

checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.from_scratch.hdf5',
verbose=1, save_best_only=True)

model.fit(train_tensors, train_targets,
validation_data=(valid_tensors, valid_targets),
epochs=epochs, batch_size=15, callbacks=[checkpointer, EarlyStopping(patience=2)], verbose=2)

Train on 6680 samples, validate on 835 samples
Epoch 1/20
Epoch 00000: val_loss improved from inf to 4.86880, saving model to saved_models/weights.best.from_scratch.hdf5
22s - loss: 4.8837 - acc: 0.0091 - val_loss: 4.8688 - val_acc: 0.0108
Epoch 2/20
Epoch 00001: val_loss improved from 4.86880 to 4.86153, saving model to saved_models/weights.best.from_scratch.hdf5
22s - loss: 4.8654 - acc: 0.0120 - val_loss: 4.8615 - val_acc: 0.0120
Epoch 3/20
Epoch 00002: val_loss improved from 4.86153 to 4.83518, saving model to saved_models/weights.best.from_scratch.hdf5
22s - loss: 4.8429 - acc: 0.0148 - val_loss: 4.8352 - val_acc: 0.0132
Epoch 4/20
Epoch 00003: val_loss improved from 4.83518 to 4.79944, saving model to saved_models/weights.best.from_scratch.hdf5
22s - loss: 4.8063 - acc: 0.0180 - val_loss: 4.7994 - val_acc: 0.0192
Epoch 5/20
Epoch 00004: val_loss improved from 4.79944 to 4.77760, saving model to saved_models/weights.best.from_scratch.hdf5
22s - loss: 4.7693 - acc: 0.0201 - val_loss: 4.7776 - val_acc: 0.0204

-- Trimmed --

Out[16]:
<keras.callbacks.History at 0x7fb34c67eb38>

### Load the Model with the Best Validation Loss¶

In [17]:
model.load_weights('saved_models/weights.best.from_scratch.hdf5')


### Test the Model¶

Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 1%.

In [18]:
# get index of predicted dog breed for each image in test set
dog_breed_predictions = [np.argmax(model.predict(np.expand_dims(tensor, axis=0))) for tensor in test_tensors]

# report test accuracy
test_accuracy = 100*np.sum(np.array(dog_breed_predictions)==np.argmax(test_targets, axis=1))/len(dog_breed_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)

Test accuracy: 6.9378%


Answer: Our accuracy of our custom CNN meets our goal "> 1%" at 7%, but we can do better.

## Step 4: Use a CNN to Classify Dog Breeds¶

To reduce training time without sacrificing accuracy, we show you how to train a CNN using transfer learning. In the following step, you will get a chance to use transfer learning to train your own CNN.

### Obtain Bottleneck Features¶

In [19]:
bottleneck_features = np.load('bottleneck_features/DogVGG16Data.npz')
train_VGG16 = bottleneck_features['train']
valid_VGG16 = bottleneck_features['valid']
test_VGG16 = bottleneck_features['test']


### Model Architecture¶

The model uses the the pre-trained VGG-16 model as a fixed feature extractor, where the last convolutional output of VGG-16 is fed as input to our model. We only add a global average pooling layer and a fully connected layer, where the latter contains one node for each dog category and is equipped with a softmax.

In [20]:
VGG16_model = Sequential()

VGG16_model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
global_average_pooling2d_2 ( (None, 512)               0
_________________________________________________________________
dense_2 (Dense)              (None, 133)               68229
=================================================================
Total params: 68,229
Trainable params: 68,229
Non-trainable params: 0
_________________________________________________________________


### Compile the Model¶

In [21]:
VGG16_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])


### Train the Model¶

In [22]:
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.VGG16.hdf5',
verbose=1, save_best_only=True)

VGG16_model.fit(train_VGG16, train_targets,
validation_data=(valid_VGG16, valid_targets),
epochs=20, batch_size=20, callbacks=[checkpointer,early_stop], verbose=2)

Train on 6680 samples, validate on 835 samples
Epoch 1/20
Epoch 00000: val_loss improved from inf to 10.71275, saving model to saved_models/weights.best.VGG16.hdf5
1s - loss: 12.0448 - acc: 0.1322 - val_loss: 10.7128 - val_acc: 0.2335
Epoch 2/20
Epoch 00001: val_loss improved from 10.71275 to 10.34588, saving model to saved_models/weights.best.VGG16.hdf5
1s - loss: 10.1500 - acc: 0.2858 - val_loss: 10.3459 - val_acc: 0.2743
Epoch 3/20
Epoch 00002: val_loss improved from 10.34588 to 10.22324, saving model to saved_models/weights.best.VGG16.hdf5
1s - loss: 9.7874 - acc: 0.3398 - val_loss: 10.2232 - val_acc: 0.2922
Epoch 4/20
Epoch 00003: val_loss improved from 10.22324 to 9.99767, saving model to saved_models/weights.best.VGG16.hdf5
1s - loss: 9.6344 - acc: 0.3666 - val_loss: 9.9977 - val_acc: 0.3090
Epoch 5/20
Epoch 00004: val_loss improved from 9.99767 to 9.85891, saving model to saved_models/weights.best.VGG16.hdf5
1s - loss: 9.4621 - acc: 0.3808 - val_loss: 9.8589 - val_acc: 0.3269

-- Trimmed --

Out[22]:
<keras.callbacks.History at 0x7fb2f47e9cc0>

### Load the Model with the Best Validation Loss¶

In [23]:
VGG16_model.load_weights('saved_models/weights.best.VGG16.hdf5')


### Test the Model¶

Now, we can use the CNN to test how well it identifies breed within our test dataset of dog images. We print the test accuracy below.

In [24]:
# get index of predicted dog breed for each image in test set
VGG16_predictions = [np.argmax(VGG16_model.predict(np.expand_dims(feature, axis=0))) for feature in test_VGG16]

# report test accuracy
test_accuracy = 100*np.sum(np.array(VGG16_predictions)==np.argmax(test_targets, axis=1))/len(VGG16_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)

Test accuracy: 42.3445%


Answer: Our Accuracy on the VGG16 dogs is doing better at 42%

### Predict Dog Breed with the Model¶

In [25]:
from extract_bottleneck_features import *

def VGG16_predict_breed(img_path):
# extract bottleneck features
bottleneck_feature = extract_VGG16(path_to_tensor(img_path))
# obtain predicted vector
predicted_vector = VGG16_model.predict(bottleneck_feature)
# return dog breed that is predicted by the model
return dog_names[np.argmax(predicted_vector)]


## Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)¶

You will now use transfer learning to create a CNN that can identify dog breed from images. Your CNN must attain at least 60% accuracy on the test set.

In Step 4, we used transfer learning to create a CNN using VGG-16 bottleneck features. In this section, you must use the bottleneck features from a different pre-trained model. To make things easier for you, we have pre-computed the features for all of the networks that are currently available in Keras:

The files are encoded as such:

Dog{network}Data.npz



where {network}, in the above filename, can be one of VGG19, Resnet50, InceptionV3, or Xception. Pick one of the above architectures, download the corresponding bottleneck features, and store the downloaded file in the bottleneck_features/ folder in the repository.

### (IMPLEMENTATION) Obtain Bottleneck Features¶

In the code block below, extract the bottleneck features corresponding to the train, test, and validation sets by running the following:

bottleneck_features = np.load('bottleneck_features/Dog{network}Data.npz')
train_{network} = bottleneck_features['train']
valid_{network} = bottleneck_features['valid']
test_{network} = bottleneck_features['test']
In [52]:
### TODO: Obtain bottleneck features from another pre-trained CNN.

# Lets use VGG19
train_VGG19 = bottleneck_features['train']
valid_VGG19 = bottleneck_features['valid']
test_VGG19 = bottleneck_features['test']


### (IMPLEMENTATION) Model Architecture¶

Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:

    <your model's name>.summary()



Question 5: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. Describe why you think the architecture is suitable for the current problem.

Answer:: My first stop in designing the architecture was to use the KerasClassifier wrapper class for use with sklearn's GridSearch hyperparameter search class. By using a programmatic approach , we can explore different values for different parameters to our model and let the machine report back which one gave the best accuracy.

We start by building the checkpointer that will save our weights if our val_loss improves. We then create a list of optimizers to pass into the GridSearch along with our model, and train every optimizer on 20 epochs. The EarlyStopping class smartly does not stop in our case, as that would halt the training pipeline.

The final CNN architecture was arrived at after much trial and error. We start with a Sequential model and add a Global Average Pooling layer for use throughout the training. We then add a Dense layer with 500 fully connected nodes which connects to our input layer of 224 nodes, and a large dropout layer that will only activate 50% of the nodes at a given time.

We add another dense layer, this time with 1000 connected nodes, and another dropout layer this time only activating 25% of the time. Finally we add one more fully connected dense layer of 113 which we use softmax to output probabilities for our 113 dog breeds.

This architecture was largely trial and error, and as I continue to learn I'd like to start documenting which neural network models worked best for each problem to gain a better intuition as to what works and what doesn't.

More details below in the comments.

In [27]:
# Here we are using a KerasClassifier with a custom function create_model
# which will run 20 epochs for each optimizer to find the one with the
# best accuracy on the training set.

# Save the best weights from our GridSearch of parameters
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.VGG19.hdf5',
verbose=1, save_best_only=True)

# Here we use a sklearn wrapper for use in our GridSearch class, which will
# do a programmatic search for the best parameters for our model
model = KerasClassifier(build_fn=create_model,epochs=20, batch_size=150, verbose=2)

shapes=[train_VGG19.shape[1:]]

# This is our parameter grid
param_grid = dict(optimizer=optimizer,shape =shapes)
# Our grid search instance, train it on VGG19 data
# EarlyStoping smartly does nothing here, since we are running a hyper parameter search
grid = GridSearchCV(estimator=model, param_grid=param_grid,fit_params={'callbacks':[checkpointer,early_stop] ,'validation_data' : (valid_VGG19, valid_targets)})
grid.fit(train_VGG19, train_targets)

## Model Architecture

# This final architecture was gotten to after much trial and error.
# Next time I would like to explore using more exhaustive with something like http://machinelearningmastery.com/grid-search-hyperparameters-deep-learning-models-python-keras/
# Except for model layer design.

# Start with a sequential model with a Global Average Pooling layer with use throughout
model = Sequential()
# Now a densely connected layer of 500, connected to our input of 224
# Add a large dropout layer , activating only 50% of them
# Add another densely connected layer, again this was largely trial and error with what gave the best performance
# A smaller dropout layer
# And our final output layer of the 133 dog breeds
model.compile(loss='categorical_crossentropy', optimizer=optimizer, metrics=['accuracy'])

Train on 4453 samples, validate on 835 samples
Epoch 1/20
Epoch 00000: val_loss improved from inf to 3.74508, saving model to saved_models/weights.best.VGG19.hdf5
0s - loss: 8.3025 - acc: 0.0359 - val_loss: 3.7451 - val_acc: 0.1940
Epoch 2/20
Epoch 00001: val_loss improved from 3.74508 to 3.16242, saving model to saved_models/weights.best.VGG19.hdf5
0s - loss: 4.4806 - acc: 0.1076 - val_loss: 3.1624 - val_acc: 0.3701
Epoch 3/20
Epoch 00002: val_loss improved from 3.16242 to 2.67736, saving model to saved_models/weights.best.VGG19.hdf5
0s - loss: 3.6776 - acc: 0.1965 - val_loss: 2.6774 - val_acc: 0.4383
Epoch 4/20
Epoch 00003: val_loss improved from 2.67736 to 2.28781, saving model to saved_models/weights.best.VGG19.hdf5
0s - loss: 3.2031 - acc: 0.2762 - val_loss: 2.2878 - val_acc: 0.4994
Epoch 5/20
Epoch 00004: val_loss improved from 2.28781 to 1.97787, saving model to saved_models/weights.best.VGG19.hdf5
0s - loss: 2.8423 - acc: 0.3342 - val_loss: 1.9779 - val_acc: 0.5449
-- Trimmed --

Out[27]:

In [47]:
from pprint import pprint
pprint(grid.grid_scores_)

[mean: 0.74820, std: 0.00441, params: {'optimizer': 'sgd', 'shape': (7, 7, 512)},
mean: 0.75913, std: 0.01086, params: {'optimizer': 'rmsprop', 'shape': (7, 7, 512)},
mean: 0.77590, std: 0.01344, params: {'optimizer': 'adam', 'shape': (7, 7, 512)},
mean: 0.78892, std: 0.00659, params: {'optimizer': 'adamax', 'shape': (7, 7, 512)},
mean: 0.75225, std: 0.00366, params: {'optimizer': 'nadam', 'shape': (7, 7, 512)}]

/home/aind2/anaconda3/envs/aind2/lib/python3.6/site-packages/sklearn/model_selection/_search.py:667: DeprecationWarning: The grid_scores_ attribute was deprecated in version 0.18 in favor of the more elaborate cv_results_ attribute. The grid_scores_ attribute will not be available from 0.20
DeprecationWarning)


Answer: We can see that our Grid Search returned ADAMAX with our greatest average score of ~79%

In [63]:
D =  { 'sgd' : 0.74820,
'rmsprop' : 0.75913,

values = [x * 100 for x in D.values() ]
plt.bar(range(len(D)), values , align='center');
plt.xticks(range(len(D)), D.keys());


### (IMPLEMENTATION) Compile the Model¶

In [69]:
# FOR FAST Training , use our recorded best parameters, just adamax for now
VGG19_model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
global_average_pooling2d_2 ( (None, 512)               0
_________________________________________________________________
dense_4 (Dense)              (None, 500)               256500
_________________________________________________________________
dropout_3 (Dropout)          (None, 500)               0
_________________________________________________________________
dense_5 (Dense)              (None, 1000)              501000
_________________________________________________________________
dropout_4 (Dropout)          (None, 1000)              0
_________________________________________________________________
dense_6 (Dense)              (None, 133)               133133
=================================================================
Total params: 890,633
Trainable params: 890,633
Non-trainable params: 0
_________________________________________________________________


### (IMPLEMENTATION) Train the Model¶

Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.

You are welcome to augment the training data, but this is not a requirement.

In [ ]:
VGG19_model.fit(train_VGG19, train_targets,
validation_data=(valid_VGG19, valid_targets),
epochs=150, batch_size=50, callbacks=[checkpointer,early_stop], verbose=2)


### (IMPLEMENTATION) Load the Model with the Best Validation Loss¶

In [ ]:
VGG19_model.load_weights('saved_models/weights.best.VGG19.hdf5')


### (IMPLEMENTATION) Test the Model¶

Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 60%.

In [7]:
### TODO: Calculate classification accuracy on the test dataset.

# get index of predicted dog breed for each image in test set
VGG19_predictions = [np.argmax(VGG19_model.predict(np.expand_dims(feature, axis=0))) for feature in test_VGG19]

# report test accuracy
test_accuracy = 100*np.sum(np.array(VGG19_predictions)==np.argmax(test_targets, axis=1))/len(VGG19_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)

Test accuracy: 78.1100%


Answer: As we can see our test accuracy is ~78%

### (IMPLEMENTATION) Predict Dog Breed with the Model¶

Write a function that takes an image path as input and returns the dog breed (Affenpinscher, Afghan_hound, etc) that is predicted by your model.

Similar to the analogous function in Step 5, your function should have three steps:

1. Extract the bottleneck features corresponding to the chosen CNN model.
2. Supply the bottleneck features as input to the model to return the predicted vector. Note that the argmax of this prediction vector gives the index of the predicted dog breed.
3. Use the dog_names array defined in Step 0 of this notebook to return the corresponding breed.

The functions to extract the bottleneck features can be found in extract_bottleneck_features.py, and they have been imported in an earlier code cell. To obtain the bottleneck features corresponding to your chosen CNN architecture, you need to use the function

extract_{network}



where {network}, in the above filename, should be one of VGG19, Resnet50, InceptionV3, or Xception.

In [70]:
import os
import random
from random import shuffle
from IPython.core.display import display, HTML

def predict_dog_breed(imgpath):
tensor = extract_VGG19(path_to_tensor(imgpath))
y_hat = VGG19_model.predict(tensor)
return dog_names[np.argmax(y_hat)]

def show_random_predictions(max_to_show):
paths = []
for root, dirs, _ in os.walk("{}/dogImages/".format(os.getcwd()), topdown=False):
for name in dirs:
path = os.path.join(root, name)
for r, dirs, files in os.walk(path, topdown=False):
i = 0
for f in files:
i += 1
if i % 5 == 0:
break
file_path = os.path.join(r,f)
paths.append(file_path)
#label = predict_dog_breed(file_path)
#print("Prediction for {} = {}".format(f, label))

shuffle(paths)
ctr = 0
for f in paths:
ctr += 1
if ctr > max_to_show:
break
label = predict_dog_breed(f)
file_name = os.path.basename(f)
if label in f:
display(HTML(("<span style='color: green'>CORRECT</span> -> Prediction of {} was {}".format(file_name, label))))
else:
display(HTML(("<span style='color: red'>INCORRECT</span>  prediction of {} was ({})".format(file_name, label))))
display_img(f)


In [66]:
show_random_predictions(10)

CORRECT -> Prediction of Bull_terrier_02788.jpg was Bull_terrier
CORRECT -> Prediction of Greater_swiss_mountain_dog_05469.jpg was Greater_swiss_mountain_dog
CORRECT -> Prediction of Norwich_terrier_07271.jpg was Norwich_terrier
CORRECT -> Prediction of Basenji_00987.jpg was Basenji