StrikingLoo

import numpy as np
np_im = pixels_from_path(file_path) #load the image as numpy matrix
pixels = vector_of_pixels(np_im) #turn it into a single vector
np.unique(pixels).shape # count unique pixel values

StrikingLoo

from dask_ml.cluster import KMeans
from PIL import Image
import numpy as np

#given the fitted k-means model and a vector of pixels, returns the new value for each pixel.
def clustered_pixels(x_fit, pixels):
    labels = x_fit.predict(pixels).compute()
    res= x_fit.cluster_centers_[labels] 
    return res

StrikingLoo

from PIL import Image
import numpy as np

def pixels_from_path(file_path):
    im = Image.open(file_path)
    np_im = np.array(im)
    #matrix of pixel RGB values
    return np_im
  
def vector_of_pixels(np_im):

StrikingLoo

print('# Fit model on training data')

history = model.fit(x_train, 
                    x_train, #we pass it the same input data as desired output
                    
                    #If the model is taking forever to train, make this bigger
                    #If it is taking forever to load for the first epoch, make this smaller
                    batch_size=256, 
                    
                    epochs=10,

StrikingLoo

customAdam = keras.optimizers.Adam(lr=0.001) #you have no idea how many times I changed this number
model.compile(optimizer=customAdam,  # Optimizer
              # Loss function to minimize
              loss="mean_squared_error",
              # List of metrics to monitor
              metrics=["mean_squared_error"])

StrikingLoo

from tensorflow import keras
from tensorflow.keras import layers

total_pixels = img_size * img_size * 3
translator_factor = 2
translator_layer_size = int(total_pixels/translator_factor)
middle_factor = 2
middle_layer_size = int(translator_layer_size/middle_factor)

inputs = keras.Input(shape=(img_size,img_size,3), name='cat_image')

StrikingLoo

from tensorflow import keras
from tensorflow.keras import layers

total_pixels = img_size * img_size * 3
translator_factor = 2
translator_layer_size = int(total_pixels/translator_factor)
middle_factor = 2
middle_layer_size = int(translator_layer_size/middle_factor)

inputs = keras.Input(shape=(img_size,img_size,3), name='cat_image')

StrikingLoo

# 10% of the data will automatically be used for validation
validation_size = 0.1
img_size = 48 # resize images to be 48x48
num_channels = 3 # RGB
sample_size = 8192 #We'll use 8192 pictures (2**13)

data = dataset.read_train_sets(train_path, img_size, ['cats'],
                               validation_size=validation_size, 
                               sample_size=sample_size)

StrikingLoo

param = {'max_depth':5,
         'eta':.1, 
         'colsample_bytree':.75 
        }
num_round = 40
new_tree = xgb.train(param, dtrain, num_round)
# make prediction
new_preds = new_tree.predict(dtest)

StrikingLoo

import math 

def msesqrt(std, test_label, preds):
    squared_errors = [diff*diff for diff in (test_label - preds)]
    return math.sqrt(sum(squared_errors)/len(preds))

msesqrt(std, test_label, first_preds) # 4.02