MODEL_DIR = "/models"
model_prefix = "ece408"
dataset_size = float("inf")
if len(sys.argv) > 1:
dataset_size = int(sys.argv[1])
if len(sys.argv) > 2:
print "Usage:", sys.argv[0], "<dataset size>"
print " <dataset_size> = [0 - 10000]"
sys.exit(-1)
# Log to stdout for MXNet
logging.getLogger().setLevel(logging.DEBUG) # logging to stdout
print "Loading fashion-mnist data...",
test_images, test_labels = load_mnist(
path="/fashion-mnist", rows=70, cols=70, kind="t10k-70")
print "done"
# Reduce the size of the dataset, if desired
dataset_size = max(0, min(dataset_size, 10000))
test_images = test_images[:dataset_size]
test_labels = test_labels[:dataset_size]
# Cap batch size at the size of our training data
batch_size = len(test_images)
# Get iterators that cover the dataset
test_iter = mx.io.NDArrayIter(
test_images, test_labels, batch_size)
# Evaluate the network
#!/usr/bin/env python
import mxnet as mx
import logging
from reader import load_mnist
# Log to stdout for MXNet
logging.getLogger().setLevel(logging.DEBUG) # logging to stdout
print "Loading fashion-mnist data...",
test_images, test_labels = load_mnist(path="/fashion-mnist", kind="t10k")
# Reshape the data to the format expected by MXNet's default convolutional layers
test_images = test_images.reshape((10000, 1, 28, 28))
test_labels = test_labels.reshape(10000)
# You can reduce the size of the train or test datasets by uncommenting the following lines
# test_images = test_images[:1000]
# test_labels = test_labels[:1000]
print "done"
# Do everything in a single batch
batch_size = len(test_images)
# Get iterators that cover the dataset
test_iter = mx.io.NDArrayIter(test_images, test_labels, batch_size)
# Evaluate the network
print "Loading model...",
lenet_model = mx.mod.Module.load(prefix='/models/baseline',
epoch=1,
context=mx.gpu())
lenet_model.bind(data_shapes=test_iter.provide_data,
import reader
import tensorflow as tf
import keras
import matplotlib.pyplot as plt
import os
# prepare data
from visible import plot_image, plot_value_array
train_img, train_label = reader.load_mnist(
'datasets', kind='train') # (60000,784) ,(60000, )
test_img, test_label = reader.load_mnist(
'datasets', kind='t10k') # (10000,784) , (10000, )
"""show img[0] in colors
plt.figure()
plt.imshow(train_img[0].reshape((28,28)))
plt.colorbar()
plt.grid(False)
plt.show()"""
# transfrom into greyscale
train_img = train_img / 255.0
test_img = test_img / 255.0
"""show img [0..24] in grey-level pic
plt.figure(figsize=(10,10))
for i in range(25):
plt.subplot(5,5,i+1)
plt.xticks([])
plt.yticks([])
plt.grid(False)
plt.imshow(train_img[i].reshape((28,28)), cmap=plt.cm.binary)
model_prefix = "ece408"
dataset_size = float("inf")
if len(sys.argv) > 1:
dataset_size = int(sys.argv[1])
if len(sys.argv) > 2:
print "Usage:", sys.argv[0], "<dataset size>"
print " <dataset_size> = [0 - 10000]"
sys.exit(-1)
# Log to stdout for MXNet
logging.getLogger().setLevel(logging.DEBUG) # logging to stdout
print "Loading fashion-mnist data..."
test_images, test_labels = load_mnist(path="/fashion-mnist",
rows=48,
cols=48,
kind="t10k-48")
print "done"
# Reduce the size of the dataset, if desired
dataset_size = max(0, min(dataset_size, 10000))
test_images = test_images[:dataset_size]
test_labels = test_labels[:dataset_size]
# Cap batch size at the size of our training data
batch_size = len(test_images)
# Get iterators that cover the dataset
test_iter = mx.io.NDArrayIter(test_images, test_labels, batch_size)
# Evaluate the network
MODEL_DIR = "/models"
model_prefix = "ece408"
dataset_size = float("inf")
if len(sys.argv) > 1:
dataset_size = int(sys.argv[1])
if len(sys.argv) > 2:
print "Usage:", sys.argv[0], "<dataset size>"
print " <dataset_size> = [0 - 10000]"
sys.exit(-1)
# Log to stdout for MXNet
logging.getLogger().setLevel(logging.DEBUG) # logging to stdout
print "Loading fashion-mnist data..."
test_images, test_labels = load_mnist(
path="/fashion-mnist", rows=64, cols=64, kind="t10k-64")
print "done"
# Reduce the size of the dataset, if desired
# dataset_size = max(0, min(dataset_size, 10000))
dataset_size = max(0, min(dataset_size, 10000))
test_images = test_images[:dataset_size]
test_labels = test_labels[:dataset_size]
# Cap batch size at the size of our training data
batch_size = len(test_images)
# Get iterators that cover the dataset
test_iter = mx.io.NDArrayIter(
test_images, test_labels, batch_size)
model_prefix = "eecs498"
dataset_size = float("inf")
if len(sys.argv) > 1:
dataset_size = int(sys.argv[1])
if len(sys.argv) > 2:
print "Usage:", sys.argv[0], "<dataset size>"
print " <dataset_size> = [0 - 10000]"
sys.exit(-1)
# Log to stdout for MXNet
logging.getLogger().setLevel(logging.DEBUG) # logging to stdout
print "Loading fashion-mnist data...",
test_images, test_labels = load_mnist(path="fashion-mnist",
rows=72,
cols=72,
kind="t10k-72")
print "done"
# Reduce the size of the dataset, if desired
dataset_size = max(0, min(dataset_size, 10000))
test_images = test_images[:dataset_size]
test_labels = test_labels[:dataset_size]
# Cap batch size at the size of our training data
batch_size = len(test_images)
# Get iterators that cover the dataset
test_iter = mx.io.NDArrayIter(test_images, test_labels, batch_size)
# Evaluate the network
return int(sum(y_predict == y)) / y.shape[0]
def get_accuraccy(self, X, y):
a, z = self.forward_propagation(X)
y_predict = self.predict(a[-1][:, 1:])
return int(sum(y_predict == y)) / y.shape[0]
def load(self, a):
fin = open(a, "rb")
self.Theta = pickle.load(fin)
print(pickle.load(fin))
fin.close()
if __name__ == '__main__':
# read data
X_train, y_train = reader.load_mnist('data/number', kind='train')
X_test, y_test = reader.load_mnist('data/number', kind='t10k')
# Normalize
mean = np.mean(X_train)
std = np.std(X_train)
X_test = (X_test - mean) / std
X_train = (X_train - mean) / std
X_train = X_train[:5000]
y_train = y_train[:5000]
X_cv = X_test[5000:10000]
y_cv = y_test[5000:10000]
X_test = X_test[:5000]
y_test = y_test[:5000]
"""Filter number 0 and 1"""
X = []
y = []
for i in range(X_test.shape[0]):
if y_test[i][0] == 0 or y_test[i][0] == 1:
X.append(X_test[i])
y.append(y_test[i])
X = np.array(X)
y = np.array(y)
return X, y
if __name__ == "__main__":
# Read data
(X_train, y_train) = reader.load_mnist('data', kind='train')
(X_test, y_test) = reader.load_mnist('data', kind='t10k')
X_train = X_train[:10000]
y_train = y_train[:10000]
X_test = X_test
y_test = y_test
# X_train, y_train = filter_array(X_train, y_train)
# X_test, y_test = filter_array(X_test, y_test)
ex1b = LogisticRegression(X_train, y_train)
y_guess = ex1b.predict(X_test)
# Get accuracy
sum = 0
for i in range(y_test.shape[0]):
