Traffic recognition.rar

#coding:utf-8 import os import random import skimage.data import skimage.transform import matplotlib import matplotlib.pyplot as plt import numpy as np import tensorflow as tf import time import argparse import sys import tempfile from skimage import color from PIL import Image print (__doc__) EPOCH_SIZE = 500 BATCH_SIZE = 40 IMAGE_SIZE =32 HIDDEN_SIZE = 256 CLASS_LABELS = 62 CROP_SIZE = 32 LEARNING_RATE = tf.Variable(0.0001, dtype=tf.float32) train_data_dir = r"C:\Users\metro_miccall\eclipse-workspace\CNN_tensorflow\CNN_traffic\datasets\BelgiumTS\Training" test_data_dir = r"C:\Users\metro_miccall\eclipse-workspace\CNN_tensorflow\CNN_traffic\datasets\BelgiumTS\Testing" #参数概要 def variable_summaries(var): with tf.name_scope('summaries'): mean = tf.reduce_mean(var) tf.summary.scalar('mean', mean)#平均值 with tf.name_scope('stddev'): stddev = tf.sqrt(tf.reduce_mean(tf.square(var - mean))) tf.summary.scalar('stddev', stddev)#标准差 tf.summary.scalar('max', tf.reduce_max(var))#最大值 tf.summary.scalar('min', tf.reduce_min(var))#最小值 tf.summary.histogram('histogram', var)#直方图 def load_datasets(data_dir,img_crop_size): directories = [d for d in os.listdir(data_dir) if os.path.isdir(os.path.join(data_dir, d))] #print("load_datasets directories:%s"%(directories)) images = [] labels = [] for d in directories: label_dir = os.path.join(data_dir, d) file_names = [os.path.join(label_dir, f) for f in os.listdir(label_dir) if f.endswith(".ppm")] for f in file_names: img = skimage.data.imread(f) image = skimage.transform.resize(img, (img_crop_size,img_crop_size )) image=color.rgb2gray(image) images.append(image) id = int(d) #print("\nid:%s"%id) labels.append(id) return images, labels def load_datasets_norm(data_dir,img_crop_size): directories = [d for d in os.listdir(data_dir) if os.path.isdir(os.path.join(data_dir, d))] #print("load_datasets directories:%s"%(directories)) images = [] labels = [] for d in directories: label_dir = os.path.join(data_dir, d) file_names = [os.path.join(label_dir, f) for f in os.listdir(label_dir) if f.endswith(".ppm")] for f in file_names: img = skimage.data.imread(f) image = skimage.transform.resize(img, (img_crop_size,img_crop_size)) image=color.rgb2gray(image) images.append(image) id = int(d) #print("\nid:%s"%id) label =[0]*CLASS_LABELS for i in range(CLASS_LABELS): if i == id: label[i] = 1 labels.append(label) images_a = np.array(images) images = images_a.reshape([-1,IMAGE_SIZE,IMAGE_SIZE,1]) labels_a = np.array(labels) labels = labels_a.reshape([-1,CLASS_LABELS]) print("\nload_datasets_norm images shape:", images.shape) print("\nload_datasets_norm labels shape:", labels.shape) return images, labels def display_images_and_labels(images, labels): #print("\ndisplay_images_and_labels labels:",labels) unique_labels = set(labels) print("\ndisplay_images_and_labels unique_labels:%s"%(unique_labels)) plt.figure(figsize=(15, 15)) i = 1 for label in unique_labels: image = images[labels.index(label)] plt.subplot(8, 8, i) plt.axis('off') plt.title("Label {0} ({1})".format(label, labels.count(label))) i += 1 _ = plt.imshow(image) plt.show() def predict_testimages(data_dir,sess,predicted_labels,images_ph,imagesize=IMAGE_SIZE): # Load the test dataset. pred_labels =[] test_images, test_labels = load_datasets(data_dir,CROP_SIZE) test_images_a = np.array(test_images) test_images_a = test_images_a.reshape([-1,IMAGE_SIZE,IMAGE_SIZE,1]) display_images_and_labels(test_images, test_labels) # Run predictions against the full test set. prediction = sess.run([predicted_labels],feed_dict={images_ph: test_images_a}) #predlen = prediction.shape[1] ''' print("predict_testimages predlen:%s"%(predlen)) testlen = len(test_labels) print("predict_testimages testlen:%s"%(testlen)) for i in range(testlen): predicted = prediction[i] pred = np.argmax(predicted) pred_labels.append(pred) match_count = sum([int(y == y_) for y, y_ in zip(test_labels, pred_labels)]) test_len = len(test_labels) accuracy = match_count / test_len print("match_count: {:.3f}".format(match_count),"test_len: {:.3f}".format(test_len)) print("Accuracy: {:.3f}".format(accuracy)) print("time_duration: {:.5f}s".format(time_duration)) ''' def input_layer(): x = tf.placeholder(tf.float32, [None, IMAGE_SIZE,IMAGE_SIZE,1], name="x-input") y_ = tf.placeholder(tf.float32, [None, CLASS_LABELS], name="y-input") keep_prob = tf.placeholder(tf.float32) return x,y_,keep_prob def compute_cross_entropy(x,y): diff = -(x * tf.log(y)) cross_entropy = tf.reduce_mean(diff) return cross_entropy def train_optimizer(cross_entropy): train_opt = tf.train.AdamOptimizer(LEARNING_RATE).minimize(cross_entropy) return train_opt def correct_prediction(x,y): correct_prediction = tf.equal(tf.argmax(x, 1), tf.argmax(y, 1)) accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32)) return accuracy def cnn_model(input_x,input_y,keep_prob): weight1=tf.Variable(tf.truncated_normal(shape=[3,3,1,32],stddev=5e-2)) kernel1=tf.nn.conv2d(input_x,weight1,[1,1,1,1],padding='SAME') bias1=tf.Variable(tf.constant(0.0,shape=[32])) conv1=tf.nn.relu(tf.nn.bias_add(kernel1,bias1)) pool1=tf.nn.max_pool(conv1,ksize=[1,2,2,1],strides=[1,2,2,1],padding='SAME') weight2=tf.Variable(tf.truncated_normal(shape=[3,3,32,64],stddev=5e-2)) kernel2=tf.nn.conv2d(pool1,weight2,[1,1,1,1],padding='SAME') bias2=tf.Variable(tf.constant(0.0,shape=[64])) conv2=tf.nn.relu(tf.nn.bias_add(kernel2,bias2)) pool2=tf.nn.max_pool(conv2,ksize=[1,2,2,1],strides=[1,2,2,1],padding='SAME') pool2_flat = tf.reshape(pool2, [-1, 8*8*64]) weight3=tf.Variable(tf.truncated_normal(shape=[8*8*64,1024],stddev=0.04)) bias3=tf.Variable(tf.constant(0.1,shape=[1024])) local3=tf.nn.relu(tf.matmul(pool2_flat,weight3)+bias3) weight4=tf.Variable(tf.truncated_normal(shape=[1024,CLASS_LABELS],stddev=5e-2)) bias4=tf.Variable(tf.constant(0.1,shape=[CLASS_LABELS])) y=tf.nn.softmax(tf.matmul(local3,weight4)+bias4) cross_entropy = compute_cross_entropy(input_y,y) train_opt = train_optimizer(cross_entropy) accuracy = correct_prediction(y,input_y) return train_opt,accuracy,y def predict_testrandomimages(data_dir,sess,predicted_labels,images_ph,imagesize=IMAGE_SIZE): # Load the test dataset. images, labels = load_datasets(data_dir,CROP_SIZE) # Pick 10 random images sample_indexes = random.sa