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.
In addition to implementing code, there will be questions that you must answer which relate to the project and your implementation. Each section where you will answer a question is preceded by a ‘Question X’ header. Carefully read each question and provide thorough answers in the following text boxes that begin with ‘Answer:’. Your project submission will be evaluated based on your answers to each of the questions and the implementation you provide.
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.
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.
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 imagestrain_targets, valid_targets, test_targets - numpy arrays containing onehot-encoded classification labelsdog_names - list of string-valued dog breed names for translating labelsfrom 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
def load_dataset(path):
data = load_files(path)
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
train_files, train_targets = load_dataset('/data/dog_images/train')
valid_files, valid_targets = load_dataset('/data/dog_images/valid')
test_files, test_targets = load_dataset('/data/dog_images/test')
# load list of dog names
# changing position from 20 -> 27
dog_names = [item[27:-1] for item in sorted(glob("/data/dog_images/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.
In the code cell below, we import a dataset of human images, where the file paths are stored in the numpy array human_files.
import random
random.seed(8675309)
# load filenames in shuffled human dataset
human_files = np.array(glob("/data/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.
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.
human_files[3]
'/data/lfw/Khatol_Mohammad_Zai/Khatol_Mohammad_Zai_0001.jpg'
import cv2
import matplotlib.pyplot as plt
%matplotlib inline
# extract pre-trained face detector
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')
# load color (BGR) image
img = cv2.imread(human_files[3])
# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# find faces in image
faces = face_cascade.detectMultiScale(gray)
# 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),3)
# 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: 1
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.
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.
# returns "True" if face is detected in image stored at img_path
def face_detector(img_path):
img = cv2.imread(img_path)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray)
return len(faces) > 0
Question 1: Use the code cell below to test the performance of the face_detector function.
human_files have a detected human face?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.
Answer:
What percentage of the first 100 images in human_files have a detected human face? 100 %
What percentage of the first 100 images in dog_files have a detected human face? 11 %
human_files_short = human_files[:100]
dog_files_short = train_files[:100]
# Do NOT modify the code above this line.
percentage_humans_detected = 100 * (sum([face_detector(file) for file in human_files_short])/float(len(human_files_short)))
percentage_dogs_detected = 100 * (sum([face_detector(file) for file in dog_files_short])/float(len(dog_files_short)))
## TODO: Test the performance of the face_detector algorithm
## on the images in human_files_short and dog_files_short.
print(" performance of OpenCV2's Haar Feature-based Cascade Classifiers model on Human Images is {0} %".format(percentage_humans_detected))
print(" performance of OpenCV2's Haar Feature-based Cascade Classifiers model on Dog Images is {0} %".format(percentage_dogs_detected))
performance of OpenCV2's Haar Feature-based Cascade Classifiers model on Human Images is 100.0 %
performance of OpenCV2's Haar Feature-based Cascade Classifiers model on Dog Images is 11.0 %
import random
def a_random_image(img_array):
img_to_display = random.choice(range(0, len(img_array)))
print("image path is \n {0}".format(img_array[img_to_display]))
img_sample = cv2.imread(img_array[img_to_display])
gray = cv2.cvtColor(img_sample, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray)
print('Number of faces detected:', len(faces))
for (x,y,w,h) in faces:
# add bounding box to color image
cv2.rectangle(img_sample,(x,y),(x+w,y+h),(255,0,0),3)
cv_rgb_sample = cv2.cvtColor(img_sample, cv2.COLOR_BGR2RGB)
plt.imshow(cv_rgb_sample)
File name - /data/dog_images/train/106.Newfoundland/Newfoundland_06989.jpg detects a human in the image.
Another weird example - /data/dog_images/train/099.Lhasa_apso/Lhasa_apso_06646.jpg detects a dog chain as human.
Play with this fuction :)
print("lets see how the images of dogs are. This function projects a random image")
a_random_image(dog_files_short)
lets see how the images of dogs are. This function projects a random image
image path is
/data/dog_images/train/106.Newfoundland/Newfoundland_06989.jpg
Number of faces detected: 1
print("lets see how the images of humans are. This function projects a random image")
a_random_image(human_files_short)
lets see how the images of humans are. This function projects a random image
image path is
/data/lfw/Michelle_Pfeiffer/Michelle_Pfeiffer_0003.jpg
Number of faces detected: 1
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?
Answer: The answer depends on the utility of the face detection application. If the application is used to login into say a banking site and expects a face detected login process, The user is very well aware that they need to be static and show the face clearly to the cam and their facial features projected in the centre of the cam( User will be very frustrated if there is a false positive login). In this case, its a reasonable expectation. If we are deploying this application to identify faces in real time in a high tension border between two countries and the App’s utility is to alarm the military operations about the alien faces in the region, this app may fail to correctly identify the aliens walking along the border. There we need a more sophisticated deep learning image/object detection algorithm deployed. Here is my response on how good the Haar cascade classifiers can be.
** Limitations and advantages of Haar Cascade classifiers **
I am summarizing my understanding from the blog
We suggest the face detector from OpenCV as a potential way to detect human images in your algorithm, but you are free to explore other approaches, especially approaches that make use of deep learning :). Please use the code cell below to design and test your own face detection algorithm. If you decide to pursue this optional task, report performance on each of the datasets.
## (Optional) TODO: Report the performance of another
## face detection algorithm on the LFW dataset
### Feel free to use as many code cells as needed.
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.
from keras.applications.resnet50 import ResNet50
# define ResNet50 model
ResNet50_model = ResNet50(weights='imagenet')
Downloading data from https://github.com/fchollet/deep-learning-models/releases/download/v0.2/resnet50_weights_tf_dim_ordering_tf_kernels.h5
102858752/102853048 [==============================] - 1s 0us/step
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
The paths_to_tensor function takes a numpy array of string-valued image paths as input and returns a 4D tensor with shape
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!
from keras.preprocessing import image
from tqdm import tqdm
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)
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.
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))
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).
### 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))
Question 3: Use the code cell below to test the performance of your dog_detector function.
human_files_short have a detected dog?dog_files_short have a detected dog?Answer:
What percentage of the images in human_files_short have a detected dog? 0 %
What percentage of the images in dog_files_short have a detected dog? 100 %
### TODO: Test the performance of the dog_detector function
### on the images in human_files_short and dog_files_short.
percentage_humans_detected_resnet = 100 * (sum([dog_detector(file) for file in human_files_short])/float(len(human_files_short)))
percentage_dogs_detected_resnet = 100 * (sum([dog_detector(file) for file in dog_files_short])/float(len(dog_files_short)))
## TODO: Test the performance of the face_detector algorithm
## on the images in human_files_short and dog_files_short.
print(" Percentage of dog images predicted by Resnet50 on Human Images is {0} %".format(percentage_humans_detected_resnet))
print(" Percentage of dog images predicted by Resnet50 on Dog Images is {0} %".format(percentage_dogs_detected_resnet))
Percentage of dog images predicted by Resnet50 on Human Images is 0.0 %
Percentage of dog images predicted by Resnet50 on Dog Images is 100.0 %
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.
| Yellow Labrador | Chocolate Labrador | Black Labrador |
|---|---|---|
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!
We rescale the images by dividing every pixel in every image by 255.
from PIL import ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True
# 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 [01:28<00:00, 48.50it/s]
100%|██████████| 835/835 [00:10<00:00, 83.26it/s]
100%|██████████| 836/836 [00:09<00:00, 83.89it/s]
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.
Answer:
The CNN architecture has 3 Pair (Conv2D, MaxPooling2D) followed by two Dropout and two Dense layer pairs. The activation functions used until the final Dense layer is relu. At the final layer I have used softmax.
The Conv2D layers discovers the hierarchies of spatial patterns in the image. Each Conv2D layer increases the depth of the Image. It is increased using the filter parameter. Here at each layer I have 2 X ed the depth of the convolutional layer from the previous layer. Hyper parameter - padding is used to ensure that the conv layer’s Spatial features are maintained as the previous layer. kernel size defines side length of the filter. Stride ( not mentioned here - defaults to 1) defines how the Conv2D filter is moved along the image.
MaxPooling2D layers are used to shrinken the height and width of the image to 1/2 its size from the previous layer.The hyper parameter used here - poolsize ( and stride, which defaults to poolsize if not mentioned) are responsible for this shrinkage. As we can see the height and width reduced from 224 to 28.
The sequence of pairs of (Conv2D, MaxPooling2D) layers gradually loses the spatial data and encodes the content of the image. At the end of the 3rd Maxpooling Layer output, the 3D array will have content which can answer questions like
Once I have deduced and lost as much of the spatial info, I Dropout and flatten the 3D array and pass it on to a Dense layer. Dense layer is a fully connected layer. To reduce the number of parameters to pass on to final Dense layer, a Droput layer is added. The Dropout layer randomly omits the percentage of data to train ( I have given 40 % ) before passing it along to a dense layer.
Final layer has softmax as its activation function. It outputs a (133,1) array where each entry is a probability that the image belongs to the following category.
The gradient descent optimization method I have used is adam and the loss function I am trying to optimize is categorical crossentropy for 133 categories. The metrics I care about is accuracy
In 5 iterations of thius model (epochs), I was able to bring in accuracy of 6.58%
from keras.layers import Conv2D, MaxPooling2D, GlobalAveragePooling2D
from keras.layers import Dropout, Flatten, Dense
from keras.models import Sequential
model = Sequential()
model.add(Conv2D(filters=16, kernel_size=2, padding='same', activation='relu',
input_shape=(224,224,3)))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=32, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=64, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Dropout(0.3))
model.add(Flatten())
model.add(Dense(500, activation='relu'))
model.add(Dropout(0.4))
model.add(Dense(133, activation='softmax'))
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, 64) 8256
_________________________________________________________________
max_pooling2d_4 (MaxPooling2 (None, 28, 28, 64) 0
_________________________________________________________________
dropout_1 (Dropout) (None, 28, 28, 64) 0
_________________________________________________________________
flatten_2 (Flatten) (None, 50176) 0
_________________________________________________________________
dense_1 (Dense) (None, 500) 25088500
_________________________________________________________________
dropout_2 (Dropout) (None, 500) 0
_________________________________________________________________
dense_2 (Dense) (None, 133) 66633
=================================================================
Total params: 25,165,677
Trainable params: 25,165,677
Non-trainable params: 0
_________________________________________________________________
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
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.
from keras.callbacks import ModelCheckpoint
### TODO: specify the number of epochs that you would like to use to train the model.
epochs = 5
### 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=20, callbacks=[checkpointer], verbose=1)
Train on 6680 samples, validate on 835 samples
Epoch 1/5
6660/6680 [============================>.] - ETA: 0s - loss: 4.8975 - acc: 0.0123Epoch 00001: val_loss improved from inf to 4.81316, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 35s 5ms/step - loss: 4.8972 - acc: 0.0124 - val_loss: 4.8132 - val_acc: 0.0144
Epoch 2/5
6660/6680 [============================>.] - ETA: 0s - loss: 4.6311 - acc: 0.0288Epoch 00002: val_loss improved from 4.81316 to 4.43745, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 33s 5ms/step - loss: 4.6299 - acc: 0.0289 - val_loss: 4.4374 - val_acc: 0.0383
Epoch 3/5
6660/6680 [============================>.] - ETA: 0s - loss: 4.2996 - acc: 0.0527Epoch 00003: val_loss improved from 4.43745 to 4.20382, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 33s 5ms/step - loss: 4.2992 - acc: 0.0528 - val_loss: 4.2038 - val_acc: 0.0695
Epoch 4/5
6660/6680 [============================>.] - ETA: 0s - loss: 3.9481 - acc: 0.1050Epoch 00004: val_loss improved from 4.20382 to 4.14097, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 33s 5ms/step - loss: 3.9475 - acc: 0.1051 - val_loss: 4.1410 - val_acc: 0.0743
Epoch 5/5
6660/6680 [============================>.] - ETA: 0s - loss: 3.3893 - acc: 0.1944Epoch 00005: val_loss did not improve
6680/6680 [==============================] - 33s 5ms/step - loss: 3.3891 - acc: 0.1942 - val_loss: 4.1909 - val_acc: 0.0958
<keras.callbacks.History at 0x7f5497467e80>
model.load_weights('saved_models/weights.best.from_scratch.hdf5')
Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 1%.
# 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: 6.5789%
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.
bottleneck_features = np.load('/data/bottleneck_features/DogVGG16Data.npz')
train_VGG16 = bottleneck_features['train']
valid_VGG16 = bottleneck_features['valid']
test_VGG16 = bottleneck_features['test']
train_VGG16.shape[1:]
(7, 7, 512)
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.
VGG16_model = Sequential()
VGG16_model.add(GlobalAveragePooling2D(input_shape=train_VGG16.shape[1:]))
VGG16_model.add(Dense(133, activation='softmax'))
VGG16_model.summary()
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
global_average_pooling2d_1 ( (None, 512) 0
_________________________________________________________________
dense_3 (Dense) (None, 133) 68229
=================================================================
Total params: 68,229
Trainable params: 68,229
Non-trainable params: 0
_________________________________________________________________
VGG16_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])
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], verbose=1)
Train on 6680 samples, validate on 835 samples
Epoch 1/20
6480/6680 [============================>.] - ETA: 0s - loss: 11.8787 - acc: 0.1319Epoch 00001: val_loss improved from inf to 10.32115, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 326us/step - loss: 11.8375 - acc: 0.1350 - val_loss: 10.3212 - val_acc: 0.2180
Epoch 2/20
6500/6680 [============================>.] - ETA: 0s - loss: 9.4764 - acc: 0.3052Epoch 00002: val_loss improved from 10.32115 to 9.41785, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 267us/step - loss: 9.4618 - acc: 0.3064 - val_loss: 9.4178 - val_acc: 0.3078
Epoch 3/20
6620/6680 [============================>.] - ETA: 0s - loss: 8.7485 - acc: 0.3801Epoch 00003: val_loss improved from 9.41785 to 8.98515, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 260us/step - loss: 8.7543 - acc: 0.3796 - val_loss: 8.9851 - val_acc: 0.3521
Epoch 4/20
6600/6680 [============================>.] - ETA: 0s - loss: 8.4569 - acc: 0.4217Epoch 00004: val_loss improved from 8.98515 to 8.93239, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 261us/step - loss: 8.4616 - acc: 0.4214 - val_loss: 8.9324 - val_acc: 0.3593
Epoch 5/20
6580/6680 [============================>.] - ETA: 0s - loss: 8.3308 - acc: 0.4442Epoch 00005: val_loss improved from 8.93239 to 8.72937, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 260us/step - loss: 8.3325 - acc: 0.4439 - val_loss: 8.7294 - val_acc: 0.3796
Epoch 6/20
6600/6680 [============================>.] - ETA: 0s - loss: 8.0916 - acc: 0.4595Epoch 00006: val_loss improved from 8.72937 to 8.53527, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 261us/step - loss: 8.0853 - acc: 0.4602 - val_loss: 8.5353 - val_acc: 0.3856
Epoch 7/20
6600/6680 [============================>.] - ETA: 0s - loss: 7.8936 - acc: 0.4783Epoch 00007: val_loss improved from 8.53527 to 8.41239, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 260us/step - loss: 7.8921 - acc: 0.4786 - val_loss: 8.4124 - val_acc: 0.4048
Epoch 8/20
6600/6680 [============================>.] - ETA: 0s - loss: 7.7004 - acc: 0.4965Epoch 00008: val_loss improved from 8.41239 to 8.28675, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 261us/step - loss: 7.7086 - acc: 0.4958 - val_loss: 8.2868 - val_acc: 0.3952
Epoch 9/20
6600/6680 [============================>.] - ETA: 0s - loss: 7.5298 - acc: 0.5112Epoch 00009: val_loss improved from 8.28675 to 8.14760, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 261us/step - loss: 7.5522 - acc: 0.5097 - val_loss: 8.1476 - val_acc: 0.4287
Epoch 10/20
6620/6680 [============================>.] - ETA: 0s - loss: 7.4627 - acc: 0.5227Epoch 00010: val_loss improved from 8.14760 to 8.14043, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 261us/step - loss: 7.4609 - acc: 0.5229 - val_loss: 8.1404 - val_acc: 0.4335
Epoch 11/20
6600/6680 [============================>.] - ETA: 0s - loss: 7.4337 - acc: 0.5282Epoch 00011: val_loss improved from 8.14043 to 8.00776, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 260us/step - loss: 7.4293 - acc: 0.5286 - val_loss: 8.0078 - val_acc: 0.4359
Epoch 12/20
6600/6680 [============================>.] - ETA: 0s - loss: 7.4216 - acc: 0.5321Epoch 00012: val_loss did not improve
6680/6680 [==============================] - 2s 258us/step - loss: 7.4178 - acc: 0.5322 - val_loss: 8.0710 - val_acc: 0.4347
Epoch 13/20
6600/6680 [============================>.] - ETA: 0s - loss: 7.4042 - acc: 0.5353Epoch 00013: val_loss did not improve
6680/6680 [==============================] - 2s 259us/step - loss: 7.4074 - acc: 0.5352 - val_loss: 8.0284 - val_acc: 0.4431
Epoch 14/20
6620/6680 [============================>.] - ETA: 0s - loss: 7.3813 - acc: 0.5372Epoch 00014: val_loss did not improve
6680/6680 [==============================] - 2s 259us/step - loss: 7.3994 - acc: 0.5361 - val_loss: 8.0830 - val_acc: 0.4383
Epoch 15/20
6620/6680 [============================>.] - ETA: 0s - loss: 7.3879 - acc: 0.5373Epoch 00015: val_loss did not improve
6680/6680 [==============================] - 2s 259us/step - loss: 7.3844 - acc: 0.5376 - val_loss: 8.0756 - val_acc: 0.4359
Epoch 16/20
6620/6680 [============================>.] - ETA: 0s - loss: 7.2486 - acc: 0.5402Epoch 00016: val_loss improved from 8.00776 to 7.93127, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 259us/step - loss: 7.2535 - acc: 0.5400 - val_loss: 7.9313 - val_acc: 0.4443
Epoch 17/20
6620/6680 [============================>.] - ETA: 0s - loss: 7.1584 - acc: 0.5474Epoch 00017: val_loss improved from 7.93127 to 7.91029, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 260us/step - loss: 7.1546 - acc: 0.5476 - val_loss: 7.9103 - val_acc: 0.4431
Epoch 18/20
6620/6680 [============================>.] - ETA: 0s - loss: 7.1214 - acc: 0.5547Epoch 00018: val_loss improved from 7.91029 to 7.80271, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 260us/step - loss: 7.1250 - acc: 0.5545 - val_loss: 7.8027 - val_acc: 0.4611
Epoch 19/20
6620/6680 [============================>.] - ETA: 0s - loss: 7.1300 - acc: 0.5541Epoch 00019: val_loss did not improve
6680/6680 [==============================] - 2s 257us/step - loss: 7.1191 - acc: 0.5548 - val_loss: 7.9129 - val_acc: 0.4491
Epoch 20/20
6600/6680 [============================>.] - ETA: 0s - loss: 7.0752 - acc: 0.5552Epoch 00020: val_loss did not improve
6680/6680 [==============================] - 2s 258us/step - loss: 7.0701 - acc: 0.5555 - val_loss: 7.8412 - val_acc: 0.4395
<keras.callbacks.History at 0x7f547d8f6518>
VGG16_model.load_weights('saved_models/weights.best.VGG16.hdf5')
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.
# 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: 44.3780%
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)]
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. These are already in the workspace, at /data/bottleneck_features. If you wish to download them on a different machine, they can be found at:
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.
The above architectures are downloaded and stored for you in the /data/bottleneck_features/ folder.
This means the following will be in the /data/bottleneck_features/ folder:
DogVGG19Data.npz
DogResnet50Data.npz
DogInceptionV3Data.npz
DogXceptionData.npz
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('/data/bottleneck_features/Dog{network}Data.npz')
train_{network} = bottleneck_features['train']
valid_{network} = bottleneck_features['valid']
test_{network} = bottleneck_features['test']
### TODO: Obtain bottleneck features from another pre-trained CNN.
bottleneck_features = np.load('/data/bottleneck_features/DogXceptionData.npz')
train_Xception = bottleneck_features['train']
valid_Xception = bottleneck_features['valid']
test_Xception = bottleneck_features['test']
train_Xception.shape
(6680, 7, 7, 2048)
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:
I have chosen Xception Model. The reasoning behind it was, it used a modified version of depthwise seperable convolutional layers. I followed these two articles to understand its significance while training a deep network.
The dataset I have in hand is a small dataset (8351 images in total). Xception model has weights from Imagenet database which does have more than 120 dog breeds. So we are dealing with a similar model + small dataset. All I have done here is cut down the top layer (final layer) of the xception model and trained it to identify 133 dog categories instead of 10000 categories from imagenet.
Here is reasoning behind the current CNN model.
I have downloaded the pre-computed bottleneck features of Xception model. That is each dog image is sent to (Xception model, imagenet weights) with its top layer cut off. The resulting output for every image is (7,7,2048)
This output is introduced as input to my custom Xception_model which has a GlobalAveragePooling2D layer and a pair of Dropout and Dense Layer.
GlobalAveragePooling2D is used to totally remove the spatial data and encode the contents of the image. It also reduces the number of features that Xception_model has to learn.
The Dropout layer is used to randomly omit 40% of the hidden units out of 272517 with which it has to finally classify 133 categories.
The dense layer is a fully connected layer which outputs the probability of image being in 133 categories. The highest probability category is considered to be the predicted label. The activation function used in this layer is softmax.
The model is compiled optimize categorical cross entropy loss function and used adam as the gradient descent function. The model has accuracy as its metric.
It was blazing fast and I could achieve 85.64% as accuracy in the hold out test set with the trained model -(epoch 20)
### TODO: Define your architecture.
Xception_model = Sequential()
Xception_model.add(GlobalAveragePooling2D(input_shape=train_Xception.shape[1:]))
Xception_model.add(Dropout(0.4))
Xception_model.add(Dense(133, activation='softmax'))
Xception_model.summary()
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
global_average_pooling2d_2 ( (None, 2048) 0
_________________________________________________________________
dropout_3 (Dropout) (None, 2048) 0
_________________________________________________________________
dense_4 (Dense) (None, 133) 272517
=================================================================
Total params: 272,517
Trainable params: 272,517
Non-trainable params: 0
_________________________________________________________________
### TODO: Compile the model.
Xception_model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
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.
### TODO: Train the model.
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.Xception.hdf5',
verbose=1, save_best_only=True)
Xception_model.fit(train_Xception, train_targets,
validation_data=(valid_Xception, valid_targets),
epochs=20, batch_size=40, callbacks=[checkpointer], verbose=1)
Train on 6680 samples, validate on 835 samples
Epoch 1/20
6520/6680 [============================>.] - ETA: 0s - loss: 1.5446 - acc: 0.6540Epoch 00001: val_loss improved from inf to 0.60162, saving model to saved_models/weights.best.Xception.hdf5
6680/6680 [==============================] - 3s 398us/step - loss: 1.5219 - acc: 0.6582 - val_loss: 0.6016 - val_acc: 0.8347
Epoch 2/20
6600/6680 [============================>.] - ETA: 0s - loss: 0.4774 - acc: 0.8626Epoch 00002: val_loss improved from 0.60162 to 0.50933, saving model to saved_models/weights.best.Xception.hdf5
6680/6680 [==============================] - 2s 361us/step - loss: 0.4755 - acc: 0.8626 - val_loss: 0.5093 - val_acc: 0.8371
Epoch 3/20
6640/6680 [============================>.] - ETA: 0s - loss: 0.3492 - acc: 0.8968Epoch 00003: val_loss improved from 0.50933 to 0.47426, saving model to saved_models/weights.best.Xception.hdf5
6680/6680 [==============================] - 2s 365us/step - loss: 0.3483 - acc: 0.8972 - val_loss: 0.4743 - val_acc: 0.8443
Epoch 4/20
6520/6680 [============================>.] - ETA: 0s - loss: 0.2763 - acc: 0.9201Epoch 00004: val_loss improved from 0.47426 to 0.45544, saving model to saved_models/weights.best.Xception.hdf5
6680/6680 [==============================] - 2s 366us/step - loss: 0.2747 - acc: 0.9208 - val_loss: 0.4554 - val_acc: 0.8467
Epoch 5/20
6480/6680 [============================>.] - ETA: 0s - loss: 0.2200 - acc: 0.9383Epoch 00005: val_loss improved from 0.45544 to 0.42881, saving model to saved_models/weights.best.Xception.hdf5
6680/6680 [==============================] - 2s 364us/step - loss: 0.2207 - acc: 0.9386 - val_loss: 0.4288 - val_acc: 0.8611
Epoch 6/20
6480/6680 [============================>.] - ETA: 0s - loss: 0.1776 - acc: 0.9563Epoch 00006: val_loss did not improve
6680/6680 [==============================] - 2s 355us/step - loss: 0.1796 - acc: 0.9548 - val_loss: 0.4629 - val_acc: 0.8599
Epoch 7/20
6480/6680 [============================>.] - ETA: 0s - loss: 0.1544 - acc: 0.9597Epoch 00007: val_loss improved from 0.42881 to 0.42571, saving model to saved_models/weights.best.Xception.hdf5
6680/6680 [==============================] - 2s 349us/step - loss: 0.1552 - acc: 0.9590 - val_loss: 0.4257 - val_acc: 0.8599
Epoch 8/20
6640/6680 [============================>.] - ETA: 0s - loss: 0.1373 - acc: 0.9672Epoch 00008: val_loss did not improve
6680/6680 [==============================] - 2s 359us/step - loss: 0.1370 - acc: 0.9674 - val_loss: 0.4323 - val_acc: 0.8683
Epoch 9/20
6520/6680 [============================>.] - ETA: 0s - loss: 0.1200 - acc: 0.9718Epoch 00009: val_loss did not improve
6680/6680 [==============================] - 2s 361us/step - loss: 0.1198 - acc: 0.9717 - val_loss: 0.4443 - val_acc: 0.8611
Epoch 10/20
6640/6680 [============================>.] - ETA: 0s - loss: 0.1015 - acc: 0.9791Epoch 00010: val_loss did not improve
6680/6680 [==============================] - 2s 364us/step - loss: 0.1013 - acc: 0.9790 - val_loss: 0.4405 - val_acc: 0.8551
Epoch 11/20
6640/6680 [============================>.] - ETA: 0s - loss: 0.0925 - acc: 0.9800Epoch 00011: val_loss did not improve
6680/6680 [==============================] - 2s 352us/step - loss: 0.0929 - acc: 0.9798 - val_loss: 0.4546 - val_acc: 0.8671
Epoch 12/20
6520/6680 [============================>.] - ETA: 0s - loss: 0.0825 - acc: 0.9833Epoch 00012: val_loss did not improve
6680/6680 [==============================] - 2s 354us/step - loss: 0.0824 - acc: 0.9831 - val_loss: 0.4575 - val_acc: 0.8623
Epoch 13/20
6640/6680 [============================>.] - ETA: 0s - loss: 0.0749 - acc: 0.9836Epoch 00013: val_loss did not improve
6680/6680 [==============================] - 2s 356us/step - loss: 0.0748 - acc: 0.9837 - val_loss: 0.4602 - val_acc: 0.8611
Epoch 14/20
6640/6680 [============================>.] - ETA: 0s - loss: 0.0669 - acc: 0.9869Epoch 00014: val_loss did not improve
6680/6680 [==============================] - 2s 361us/step - loss: 0.0670 - acc: 0.9870 - val_loss: 0.4690 - val_acc: 0.8611
Epoch 15/20
6560/6680 [============================>.] - ETA: 0s - loss: 0.0629 - acc: 0.9884Epoch 00015: val_loss did not improve
6680/6680 [==============================] - 2s 357us/step - loss: 0.0635 - acc: 0.9883 - val_loss: 0.4724 - val_acc: 0.8599
Epoch 16/20
6640/6680 [============================>.] - ETA: 0s - loss: 0.0563 - acc: 0.9902Epoch 00016: val_loss did not improve
6680/6680 [==============================] - 2s 347us/step - loss: 0.0563 - acc: 0.9901 - val_loss: 0.4732 - val_acc: 0.8587
Epoch 17/20
6520/6680 [============================>.] - ETA: 0s - loss: 0.0538 - acc: 0.9899Epoch 00017: val_loss did not improve
6680/6680 [==============================] - 2s 349us/step - loss: 0.0541 - acc: 0.9898 - val_loss: 0.4810 - val_acc: 0.8587
Epoch 18/20
6600/6680 [============================>.] - ETA: 0s - loss: 0.0483 - acc: 0.9926Epoch 00018: val_loss did not improve
6680/6680 [==============================] - 2s 357us/step - loss: 0.0480 - acc: 0.9927 - val_loss: 0.4643 - val_acc: 0.8575
Epoch 19/20
6640/6680 [============================>.] - ETA: 0s - loss: 0.0445 - acc: 0.9926Epoch 00019: val_loss did not improve
6680/6680 [==============================] - 2s 359us/step - loss: 0.0448 - acc: 0.9925 - val_loss: 0.4852 - val_acc: 0.8551
Epoch 20/20
6640/6680 [============================>.] - ETA: 0s - loss: 0.0442 - acc: 0.9919Epoch 00020: val_loss did not improve
6680/6680 [==============================] - 2s 366us/step - loss: 0.0443 - acc: 0.9918 - val_loss: 0.4968 - val_acc: 0.8563
<keras.callbacks.History at 0x7f5349b55ba8>
### TODO: Load the model weights with the best validation loss.
Xception_model.load_weights('saved_models/weights.best.Xception.hdf5')
Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 60%.
### TODO: Calculate classification accuracy on the test dataset.
Xception_predictions = [np.argmax(Xception_model.predict(np.expand_dims(feature, axis=0))) for feature in test_Xception]
# report test accuracy
test_accuracy = 100*np.sum(np.array(Xception_predictions)==np.argmax(test_targets, axis=1))/len(Xception_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
Test accuracy: 85.6459%
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:
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.
### TODO: Write a function that takes a path to an image as input
### and returns the dog breed that is predicted by the model.
def Xception_predict_breed(img_path):
# extract bottleneck features
bottleneck_feature = extract_Xception(path_to_tensor(img_path))
# obtain predicted vector
predicted_vector = Xception_model.predict(bottleneck_feature)
# return dog breed that is predicted by the model
return dog_names[np.argmax(predicted_vector)]
Write an algorithm that accepts a file path to an image and first determines whether the image contains a human, dog, or neither. Then,
You are welcome to write your own functions for detecting humans and dogs in images, but feel free to use the face_detector and dog_detector functions developed above. You are required to use your CNN from Step 5 to predict dog breed.
Some sample output for our algorithm is provided below, but feel free to design your own user experience!
### TODO: Write your algorithm.
### Feel free to use as many code cells as needed.
def dog_app(img_path):
img_output = cv2.imread(img_path)
cv_rgb_output = cv2.cvtColor(img_output, cv2.COLOR_BGR2RGB)
predicted_breed = Xception_predict_breed(img_path)
if dog_detector(img_path):
print("hello, dog!")
plt.imshow(cv_rgb_output)
print(" you belong to dog breed {0}".format(predicted_breed))
elif face_detector(img_path):
print("hello, human!")
plt.imshow(cv_rgb_output)
plt.show()
print("You are very close to dog breed {0}".format(predicted_breed))
else:
plt.imshow(cv_rgb_output)
plt.show()
print("ERROR!!!! ALIENS SPOTTED")
return predicted_breed
In this section, you will take your new algorithm for a spin! What kind of dog does the algorithm think that you look like? If you have a dog, does it predict your dog’s breed accurately? If you have a cat, does it mistakenly think that your cat is a dog?
Test your algorithm at least six images on your computer. Feel free to use any images you like. Use at least two human and two dog images.
Question 6: Is the output better than you expected :) ? Or worse :( ? Provide at least three possible points of improvement for your algorithm.
Answer:
I have shared my observation with my experimentation below.
cat_images = glob("test_images/cats/*")
for each_cat in cat_images:
dog_app(each_cat)
Downloading data from https://github.com/fchollet/deep-learning-models/releases/download/v0.4/xception_weights_tf_dim_ordering_tf_kernels_notop.h5
83689472/83683744 [==============================] - 1s 0us/step
ERROR!!!! ALIENS SPOTTED
ERROR!!!! ALIENS SPOTTED
I expected this to work fine. The Restnet50 model is trained initally to check if the image falls under categories 151-268 of Imagenet database . these two images didnt pass that test for sure. So it checked with Haarcascade human face detector. It failed there as well. So our algorithm turned cats to aliens :D
human_images = glob("test_images/humans/*")
for each_human in human_images:
dog_app(each_human)
hello, human!
You are very close to dog breed Lowchen
hello, human!
You are very close to dog breed Lowchen
I am unsure why the model predicted me and Donald trump to the same breed. We don’t have similar color features (be it hair or skin color). I am smiling and I have a long due unshaven beard while DJT is clean shaven. His is a side projected image while I am facing the camera. I have furthered my research later in the section
This is a game. If you pass the image array and image target dog_app_checker will predict if the dog is actually classified to correct breed or not. if it is classified to correct breed, you will have an affirmation in Green color else you will have a warning in red color.
class bcolors:
GREEN = '\033[92m'
RED = '\033[91m'
ENDC = '\033[0m'
def dog_app_checker(img_array, img_target):
# randomly select a dog from test image
img_to_display = random.choice(range(0,len(img_array)))
predicted_breed = dog_app(img_array[img_to_display])
actual_label = dog_names[np.argmax(img_target[img_to_display])]
if predicted_breed==actual_label:
output = "SUCCESS! Actual label is {0}".format(actual_label)
print(bcolors.GREEN +output+bcolors.ENDC)
else:
output = "FAILURE! Actual label is Testing{0} and predicted breed is {1}".format(actual_label, predicted_breed)
print(bcolors.RED +output+bcolors.ENDC)
# predict the label and compare with the original test target
dog_app_checker(test_files,test_targets)
hello, dog!
you belong to dog breed Bullmastiff
[92mSUCCESS! Actual label is Bullmastiff[0m
The following section, I have shown original vs predicted plots and I tried to reason out why a prediction work or why it did not.
def read_img_helper(path):
img = cv2.imread(path)
return cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
def return_pred_and_its_file_names(file_list):
predicted_breed_list = list()
predicted_breed_sample = list()
for idx, each_image in enumerate(file_list):
prediction = Xception_predict_breed(file_list[idx])
sample_file = [x for x in train_files if prediction in x][0]
predicted_breed_list.append(prediction)
predicted_breed_sample.append(sample_file)
return predicted_breed_list, predicted_breed_sample
def plot_original_and_predicted(file_list):
fig = plt.figure(figsize=(32, 32))
rows = len(file_list)
counter = 0
predicted_breed_list, predicted_breed_sample = return_pred_and_its_file_names(file_list)
for idx, image in enumerate(file_list):
counter +=1
ax = fig.add_subplot(rows, 2, counter, xticks=[], yticks=[])
ax.set_title("original")
ax.imshow(read_img_helper(file_list[idx]))
counter +=1
ax1 = fig.add_subplot(rows, 2, counter, xticks=[], yticks=[])
ax1.set_title("predicted as {0}".format(predicted_breed_list[idx]))
ax1.imshow(read_img_helper(predicted_breed_sample[idx]))
animated_dog_files = glob("test_images/animated_dogs/*")
plot_original_and_predicted(animated_dog_files)
None of the above predictions were right. but I could empathise with the model. Also I understand that the animations are exagerrations of reality. But some striking observations here.
plot_original_and_predicted(human_images)
I am not sure how to reason the above predicitons.
My next steps to improve dog app:
Multi Object identification can be introduced to this dog app. Identifying more than one dog breed present in the image. That makes the data collection process easier. We don’t have to do any more cleansing part.
I have not done image augumented training. That is training that learns scale, rotation and transalation invariant images. I believe that passing ton the augumented image to the Xception network to produce intermediate (7 X 7 X 2048) features for each image would enrich the classification accuracy.
I would like to dabble with hyper parameters like gradient descent algorithm of choice. I have triend out Adam and RMSProp. But I would like to see how model reacts for algorithms like AdaMax, AdaGrad, nestrov Accerlated Momentum, etc.
multi_dogs = glob("test_images/dangerous/*")
plot_original_and_predicted(multi_dogs)
These are supoosedly the most dangerous dog breeds Again the model failed me with the prediction. Time to retrain :D