shayaf84

%tensorflow_version 1.x
class pkg:
  #### DOWNLOADING AND LOADING DATA
  def get_metadata(metadata_path, which_splits = ['train', 'test']):  
    
    metadata = pd.read_csv(metadata_path)
    keep_idx = metadata['split'].isin(which_splits)
    return metadata[keep_idx]

  def get_data_split(split_name, flatten, all_data, metadata, image_shape):

shayaf84

# get a table with information about ALL of our images
metadata = get_metadata()

# what does it look like?
metadata.head()

shayaf84

print(metadata.groupby(["class","split"]).count())

sns.countplot(x ="class", data=metadata, hue = "split")

shayaf84

def plot_one_image(data, labels = [], index = None, image_shape = [64,64,3]):
   
    num_dims   = len(data.shape)
    num_labels = len(labels)

    # reshape data if necessary
    if num_dims == 1:
      data = data.reshape(target_shape)
    if num_dims == 2:
      data = data.reshape(np.vstack[-1, image_shape])

shayaf84

train_data, train_labels = get_train_data()
test_data, test_labels = get_test_data()

plot_one_image(train_data,train_labels,1)
plot_one_image(train_data,train_labels,40)

shayaf84

### Here we load the train and test data for you to use.
(train_data, train_labels) = get_train_data(flatten = True)
(test_data, test_labels) = get_test_data(flatten = True)
############

knn  = KNeighborsClassifier(n_neighbors=10)
log_reg  = LogisticRegression()


knn.fit(train_data, train_labels)

shayaf84

def plot_acc(history, ax = None, xlabel = 'Epoch #'):
    history = history.history
    history.update({'epoch':list(range(len(history['val_accuracy'])))})
    history = pd.DataFrame.from_dict(history)
    best_epoch = history.sort_values(by = 'val_accuracy', \
                                     ascending = False).iloc[0]['epoch']
    if not ax:
      f, ax = plt.subplots(1,1)
    sns.lineplot(x = 'epoch', y = 'val_accuracy', data = history \
                 label = 'Validation', ax = ax)

shayaf84

model = Sequential()
model.add(Reshape((64,64,3)))
model.add(Conv2D(32, (3, 3), padding = 'same', \
                 activation="relu",input_shape = (2000,64,64,3)))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.5))
model.add(Conv2D(32, (3, 3), padding = 'same',activation="relu"))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Flatten()) 
model.add(Dropout(0.5))

shayaf84

expert_conv = VGG16(weights = 'imagenet', include_top = False,input_shape=(64,64,3))

for layer in expert_conv.layers:
      trainable = True
      layer.trainable = trainable

expert_model = Sequential()
expert_model.add(Reshape((64,64,3)))
expert_model.add(expert_conv)
expert_model.add(GlobalAveragePooling2D())

shayaf84

# Packages used
import numpy as np 
import pandas as pd 
import seaborn as sns
import matplotlib.pyplot as plt
from imblearn.over_sampling import BorderlineSMOTE
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import precision_score, recall_score, f1_score, confusion_matrix
from sklearn.linear_model import LogisticRegression