Deep CNN Models

Constructing and training your own ConvNet from scratch can be Hard and a long task.

A common trick used in Deep Learning is to use a pre-trained model and finetune it to the specific data it will be used for.

Famous Models with Keras

This notebook contains code and reference for the following Keras models (gathered from https://github.com/fchollet/keras/tree/master/keras/applications)

VGG16
VGG19
ResNet50
Inception v3
Xception
... more to come

References

Very Deep Convolutional Networks for Large-Scale Image Recognition - please cite this paper if you use the VGG models in your work.
Deep Residual Learning for Image Recognition - please cite this paper if you use the ResNet model in your work.
Rethinking the Inception Architecture for Computer Vision - please cite this paper if you use the Inception v3 model in your work.

All architectures are compatible with both TensorFlow and Theano, and upon instantiation the models will be built according to the image dimension ordering set in your Keras configuration file at ~/.keras/keras.json.

For instance, if you have set image_data_format="channels_last", then any model loaded from this repository will get built according to the TensorFlow dimension ordering convention, "Width-Height-Depth".

VGG16

VGG19

`keras.applications`



In [1]:

    
from keras.applications import VGG16
from keras.applications.imagenet_utils import preprocess_input, decode_predictions
import os









    



Using TensorFlow backend.



In [2]:

    
# -- Jupyter/IPython way to see documentation
# please focus on parameters (e.g. include top)
VGG16??



In [3]:

    
vgg16 = VGG16(include_top=True, weights='imagenet')









    



Downloading data from https://github.com/fchollet/deep-learning-models/releases/download/v0.1/vgg16_weights_tf_dim_ordering_tf_kernels.h5

If you're wondering where this HDF5 files with weights is stored, please take a look at ~/.keras/models/

HandsOn VGG16 - Pre-trained Weights



In [6]:

    
IMAGENET_FOLDER = 'imgs/imagenet'  #in the repo



In [10]:

    
!ls imgs/imagenet









    



apricot_565.jpeg     apricot_787.jpeg     strawberry_1174.jpeg
apricot_696.jpeg     strawberry_1157.jpeg strawberry_1189.jpeg



In [13]:

    
from keras.preprocessing import image
import numpy as np

img_path = os.path.join(IMAGENET_FOLDER, 'strawberry_1157.jpeg')
img = image.load_img(img_path, target_size=(224, 224))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
print('Input image shape:', x.shape)

preds = vgg16.predict(x)
print('Predicted:', decode_predictions(preds))









    



Input image shape: (1, 224, 224, 3)
Downloading data from https://s3.amazonaws.com/deep-learning-models/image-models/imagenet_class_index.json
Predicted: [[('n07745940', 'strawberry', 0.98483676), ('n07836838', 'chocolate_sauce', 0.0073711565), ('n07614500', 'ice_cream', 0.0030998574), ('n04332243', 'strainer', 0.0025101686), ('n04476259', 'tray', 0.00060175249)]]



In [14]:

    
img_path = os.path.join(IMAGENET_FOLDER, 'apricot_696.jpeg')
img = image.load_img(img_path, target_size=(224, 224))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
print('Input image shape:', x.shape)

preds = vgg16.predict(x)
print('Predicted:', decode_predictions(preds))









    



Input image shape: (1, 224, 224, 3)
Predicted: [[('n07747607', 'orange', 0.87526792), ('n07749582', 'lemon', 0.03620464), ('n07717556', 'butternut_squash', 0.021843448), ('n03937543', 'pill_bottle', 0.0126132), ('n03942813', 'ping-pong_ball', 0.0054204506)]]



In [15]:

    
img_path = os.path.join(IMAGENET_FOLDER, 'apricot_565.jpeg')
img = image.load_img(img_path, target_size=(224, 224))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
print('Input image shape:', x.shape)

preds = vgg16.predict(x)
print('Predicted:', decode_predictions(preds))









    



Input image shape: (1, 224, 224, 3)
Predicted: [[('n07718472', 'cucumber', 0.29338178), ('n07716358', 'zucchini', 0.2383192), ('n04596742', 'wok', 0.042132568), ('n07716906', 'spaghetti_squash', 0.038422), ('n07711569', 'mashed_potato', 0.036552209)]]

Hands On:

Try to do the same with VGG19 Model



In [ ]:

    
# from keras.applications import VGG19

Residual Networks

ResNet 50



In [ ]:

    
## from keras.applications import ...



In [ ]: