Skip to content

Instantly share code, notes, and snippets.

Embed
What would you like to do?
#Libraries we will need
from __future__ import division
import numpy as np
import os
from skimage import data
import matplotlib.pyplot as plt
DATA_PATH = 'traffic/'
def load_data(dir):
directories = [d for d in os.listdir(dir)
if os.path.isdir(os.path.join(dir,d))]
labels = []
images = []
for d in directories:
label_directory = os.path.join(dir, d)
file_names = [os.path.join(label_directory, f)
for f in os.listdir(label_directory)
if f.endswith(".ppm")]
for f in file_names:
images.append(data.imread(f))
labels.append(int(d))
_labels = np.array(labels)
return images,_labels
from skimage import transform
from skimage.color import rgb2gray
# Image rescaling and dataset batching
images28 = [transform.resize(image, (28, 28)) for image in images]
images28 = np.array(images28)
images28 = rgb2gray(images28)
images28_batches = np.vsplit(images28,25)
labels_batches = np.split(labels,25)
print np.shape(images28_batches)
#Simple NN architecture
#Accuracy after 3 epochs 0.505158730159
#define placeholder for our images and labels
x = tf.placeholder(dtype = tf.float32, shape = [None, 28, 28])
y = tf.placeholder(dtype = tf.int32, shape = [None])
#layer 1 -> flat an image to 1D vector
images_flat = tf.contrib.layers.flatten(x)
#Pass layer 1 answer to a fully connected layer with relu activations
logits = tf.contrib.layers.fully_connected(images_flat, 62, tf.nn.relu)
#Compute the loss
loss = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(labels = y, logits = logits))
#Minimize it with an optimizer
train_op = tf.train.AdamOptimizer(learning_rate=0.001).minimize(loss)
#Give me back the correct prediction for each training example. Remember the output is a 62 position vector
#with a probability distribution over the labels
correct_pred = tf.argmax(logits, 1)
# Compute accuracy for training
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
print("images_flat: ", images_flat)
print("logits: ", logits)
print("loss: ", loss)
print("predicted_labels: ", correct_pred)
#Convolutional architecture
#Accuracy after three epochs Accuracy: 0.922
#define placeholder for our images and labels
x = tf.placeholder(dtype = tf.float32, shape = [None, 28, 28])
y = tf.placeholder(dtype = tf.int32, shape = [None])
conv1_weights = tf.Variable(tf.truncated_normal([2, 2, 1, 32],
stddev=0.1,seed=tf.set_random_seed(1234),
dtype=tf.float32))
conv1_biases = tf.Variable(tf.zeros([32], dtype=tf.float32))
#this time we do not make our image flat, but let it as it is.
input_layer = tf.reshape(x, [-1, 28, 28, 1])
# Convolutional Layer #1
conv1 = tf.layers.conv2d(
inputs=input_layer,
filters=32,
kernel_size=[5, 5],
padding="same",
activation=tf.nn.relu)
# Pooling Layer #1
pool1 = tf.layers.max_pooling2d(inputs=conv1, pool_size=[2, 2], strides=2)
# Convolutional Layer #2 and Pooling Layer #2
conv2 = tf.layers.conv2d(
inputs=pool1,
filters=64,
kernel_size=[5, 5],
padding="same",
activation=tf.nn.relu)
pool2 = tf.layers.max_pooling2d(inputs=conv2, pool_size=[2, 2], strides=2)
# Dense Layer
pool2_flat = tf.reshape(pool2, [-1, 7 * 7 * 64])
dense = tf.layers.dense(inputs=pool2_flat, units=1024, activation=tf.nn.relu)
dropout = tf.layers.dropout(inputs=dense, rate=0.4, training=True)
# Logits Layer
logits = tf.layers.dense(inputs=dropout, units=62)
#Finally I flatten the image and pass it to a full dense layer so it can give me my probabily distribution
loss = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(labels = y, logits =logits))
train_op = tf.train.AdamOptimizer(learning_rate=0.001).minimize(loss)
correct_pred = tf.argmax(logits, 1)
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
print("images_flat: ", images_flat)
print("logits: ", logits)
print("loss: ", loss)
print("predicted_labels: ", correct_pred)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.