shgidi

from matplotlib import pyplot as plt
import numpy as np
%matplotlib inline
#show and tell

fig = plt.figure(figsize=(16,12))
n_images_per_row=4
choices=np.random.choice(len(imgs),16)
#choices=range(16)
for i,id in enumerate(choices):

shgidi

import xgboost as xgb

clf = xgb.XGBClassifier(max_depth=10, n_estimators=1500,min_child_weight=9,learning_rate=0.01,
                       nthread=8, subsample=0.80,colsample_bytree=0.80,seed=4242)


clf.fit(trn , y,eval_set=[(val, y_val)], eval_metric='mlogloss', verbose=True, early_stopping_rounds=50)
#multiclass logloss

shgidi

#genertor
train_datagen = ImageDataGenerator(
        rescale=1./255,
        shear_range=0.2,
        zoom_range=0.2,
        horizontal_flip=True) #many other augmentations are avialbile

#flow
datagen.flow(x, batch_size=1, save_to_dir='preview', save_prefix='cat', save_format='jpeg')

shgidi

import bcolz
import pickle

def save_array(fname, arr):
    c=bcolz.carray(arr, rootdir=fname, mode='w')
    c.flush()


def load_array(fname):
    return bcolz.open(fname)[:]

shgidi

plt.imshow(np.array(Image.open(g[0]))[y-30:y+30,x-30:x+30])
currentAxis = plt.gca()
coords=[10,10],20,20
currentAxis.add_patch(plt.Rectangle(*coords, fill=False, edgecolor=color, linewidth=2))

shgidi

from keras.applications.resnet50 import ResNet50
from scipy.misc import imresize

resnet=ResNet50()
imgs_new=[]

for i,img in enumerate(imgs_new_rand):
    imgs_new[i]=imresize(img,(224,224,3))
    
resnet_preds=resnet.predict(imgs_new)

shgidi

from keras.applications.vgg16 import VGG16
from keras.layers import Conv2D
from keras.models import Sequential
from keras.layers import BatchNormalization
from keras.optimizers import Adam

vgg=VGG16()

p=0.4 #dropout
label_count=17

shgidi

#simple model
model = Sequential([
    #Embedding(vocab_size, vec_size, input_length=seq_len,weights=[emb]),
    #Flatten(),
    Dense(100, input_dim=test_data.shape[1], init='uniform', activation='relu'),
    #Dropout(0.7),
    Dense(1, activation='sigmoid')])
model.compile(loss='binary_crossentropy', optimizer=Adam(), metrics=['accuracy'])

shgidi

from keras.models import model_from_json

def save_keras_model(model,path):
  model_json = model.to_json()
  with open(path+"json", "w") as json_file:
    json_file.write(model_json)
  model.save_weights(path+'.hdf5')
  
  
def load_keras_model(path):

shgidi

#https://gist.github.com/stared/dfb4dfaf6d9a8501cd1cc8b8cb806d2e
class PlotLosses(keras.callbacks.Callback):
    
    def __init__(self,imgs):
        super(PlotLosses, self).__init__()
        self.imgs=imgs
    
    def on_train_begin(self, logs={}):
        self.i = 0
        self.x = []