卷积神经网络处理Cifar-10分类数据集

2023-05-16

Cifar-10分类数据集

Cifar-10分类数据集简介

CIFAR-10数据集由10个类的60000个32x32彩色图像组成，每个类有6000个图像。有50000个训练图像和10000个测试图像。
数据集分为五个训练批次和一个测试批次，每个批次有10000个图像。测试批次包含来自每个类别的恰好1000个随机选择的图像。

数据
- 一个10000x3072二维uint8类型的numpy数组。
- 数组的每一行存储一张32x32彩色图像，即每一行存储32323=3072个数字信息。其中，前1024个元素为红色通道，随后1024个元素为绿色通道，最后1024个元素为蓝色通道。
- 图像以行优先的顺序存储，即数组的前32个元素对应图像第一行像素红色通道。
标签
- 取值范围为0-9的10000个数字组成的列表。
- 索引i处的数字表示数据数组中第i行对应图像的标签。
batches.meta文件
- python字典对象
- label_names：包含10个元素的列表，为上述标签数组中的数字赋予名称。
  例如，label_names[0]=“飞机”，label_names[1]=“汽车”等

代码

%xmode Verbose
%pdb on

Exception reporting mode: Verbose
Automatic pdb calling has been turned ON

import matplotlib.pyplot as plt
import numpy as np
import os
import pickle
import tensorflow as tf

1. 定义Cifar-10类

class CifarLoader(object):
    
    def __init__(self, source_files):
        self._source = source_files
        self._i = 0
        self.images = None
        self.labels = None
        
    def load(self):
        data = [unpickle(f) for f in self._source]
        images = np.vstack([d[b"data"] for d in data])
        n = len(images)
        self.images = images.reshape(n, 3, 32, 32).transpose(0, 2, 3, 1).astype(float) / 255
        self.labels = one_hot(np.hstack([d[b"labels"] for d in data]), 10)
        return self
    
    def next_batch(self, batch_size):
        x, y = self.images[self._i : self._i + batch_size], self.labels[self._i : self._i + batch_size]
        self._i = (self._i + batch_size) % len(self.images)
        return x, y

2. 定义函数

DATA_PATH = "../dataset/cifar-10-batches-py/"

def unpickle(file):
    with open(os.path.join(DATA_PATH, file), "rb") as fo:
        dict = pickle.load(fo, encoding="bytes")
    return dict

def one_hot(vec, vals=10):
    n = len(vec)
    out = np.zeros((n, vals))
    out[range(n), vec] = 1
    return out

3. 定义数据管理器类

class CifarDataManager(object):
    
    def __init__(self):
        self.train = CifarLoader(["data_batch_{}".format(i) for i in range(1, 6)]).load()
        self.test = CifarLoader(["test_batch"]).load()

4. 显示Cifar-10数据集图片

def display_cifar(images, size):
    n = len(images)
    plt.figure()
    plt.gca().set_axis_off()
    im = np.vstack([np.hstack([images[np.random.choice(n)] for i in range(size)]) for i in range(size)])
    plt.imshow(im)
    plt.show()

4.1 Cifar-10数据集显示

cifar = CifarDataManager()
print("Number of train images: {}".format(len(cifar.train.images)))
print("Number of train labels: {}".format(len(cifar.train.labels)))
print("Number of test images: {}".format(len(cifar.test.images)))
print("Number of test labels: {}".format(len(cifar.test.labels)))
images = cifar.train.images

display_cifar(images, 10)

Number of train images: 50000
Number of train labels: 50000
Number of test images: 10000
Number of test labels: 10000

Cifar-10

5. 神经网络

5.1 定义函数

def weight_variable(shape):
    initial = tf.truncated_normal(shape, stddev=0.1)
    return tf.Variable(initial)

def bias_variable(shape):
    initial = tf.constant(0.1, shape=shape)
    return tf.Variable(initial)

def conv2d(x, W):
    return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding="SAME")

def max_pool_2x2(x):
    return tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")

def conv_layer(input, shape):
    W = weight_variable(shape)
    b = bias_variable([shape[3]])
    return tf.nn.relu(conv2d(input, W) + b)

def full_layer(input, size):
    in_size = int(input.get_shape()[1])
    W = weight_variable([in_size, size])
    b = bias_variable([size])
    return tf.matmul(input, W) + b

5.2 网络结构

5.2.1 - 5.2.3 给出3种网络结构，准确率分别为70%、75%、83%，运行时选择其中一种结构

x = tf.placeholder(dtype=tf.float32, shape=[None, 32, 32, 3])
y_ = tf.placeholder(dtype=tf.float32, shape=[None, 10])
keep_prob = tf.placeholder(dtype=tf.float32)

5.2.1 双层卷积网络

# conv 1
conv1 = conv_layer(x, [5, 5, 3, 32])
conv1_pool = max_pool_2x2(conv1)

# conv 2
conv2 = conv_layer(conv1_pool, [5, 5, 32, 64])
conv2_pool = max_pool_2x2(conv2)
conv2_flat = tf.reshape(conv2_pool, [-1, 8 * 8 * 64])

# full 1
full1 = tf.nn.relu(full_layer(conv2_flat, 1024))
full1_drop = tf.nn.dropout(full1, keep_prob=keep_prob)

# full 2
y_conv = full_layer(full1_drop, 10)

5.2.2 三层卷积网络

# conv 1
conv1 = conv_layer(x, [5, 5, 3, 32])
conv1_pool = max_pool_2x2(conv1)

# conv 2
conv2 = conv_layer(conv1_pool, [5, 5, 32, 64])
conv2_pool = max_pool_2x2(conv2)

# conv 3
conv3 = conv_layer(conv2_pool, [5, 5, 64, 128])
conv3_pool = max_pool_2x2(conv3)
conv3_flat = tf.reshape(conv3_pool, [-1, 4 * 4 * 128])
conv3_drop = tf.nn.dropout(conv3_flat, keep_prob=keep_prob)

# full 1
full1 = tf.nn.relu(full_layer(conv3_drop, 512))
full1_drop = tf.nn.dropout(full1, keep_prob=keep_prob)

# full 2
y_conv = full_layer(full1_drop, 10)

5.2.3 多层卷积网络

C1, C2, C3 = 30, 50, 80
F1 = 500

# conv block 1
conv1_1 = conv_layer(x, shape=[3, 3, 3, C1])
conv1_2 = conv_layer(conv1_1, shape=[3, 3, C1, C1])
conv1_3 = conv_layer(conv1_2, shape=[3, 3, C1, C1])
conv1_pool = max_pool_2x2(conv1_3)
conv1_drop = tf.nn.dropout(conv1_pool, keep_prob=keep_prob)

# conv block 2
conv2_1 = conv_layer(conv1_drop, shape=[3, 3, C1, C2])
conv2_2 = conv_layer(conv2_1, shape=[3, 3, C2, C2])
conv2_3 = conv_layer(conv2_2, shape=[3, 3, C2, C2])
conv2_pool = max_pool_2x2(conv2_3)
conv2_drop = tf.nn.dropout(conv2_pool, keep_prob=keep_prob)

# conv block 3
conv3_1 = conv_layer(conv2_drop, shape=[3, 3, C2, C3])
conv3_2 = conv_layer(conv3_1, shape=[3, 3, C3, C3])
conv3_3 = conv_layer(conv3_2, shape=[3, 3, C3, C3])
conv3_pool = tf.nn.max_pool(conv3_3, ksize=[1, 8, 8, 1], strides=[1, 8, 8, 1], padding="SAME")

conv3_flat = tf.reshape(conv3_pool, [-1, C3])
conv3_drop = tf.nn.dropout(conv3_flat, keep_prob=keep_prob)

# full layer 1
full1 = tf.nn.relu(full_layer(conv3_drop, F1))
full1_drop = tf.nn.dropout(full1, keep_prob=keep_prob)

y_conv = full_layer(full1_drop, 10)

5.3 训练设定

NUM_STEPS = 50000
BATCH_SIZE = 100

# 交叉熵
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits=y_conv, labels=y_)
cross_entropy = tf.reduce_mean(cross_entropy)
train_step = tf.train.AdamOptimizer(1e-3).minimize(cross_entropy)

# 识别率
correct_prediction = tf.equal(tf.arg_max(y_conv, 1), tf.arg_max(y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

WARNING:tensorflow:From <ipython-input-11-1431c7e98006>:5: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version.
Instructions for updating:

Future major versions of TensorFlow will allow gradients to flow
into the labels input on backprop by default.

See `tf.nn.softmax_cross_entropy_with_logits_v2`.

WARNING:tensorflow:From <ipython-input-11-1431c7e98006>:10: arg_max (from tensorflow.python.ops.gen_math_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Use `argmax` instead

5.4 会话

def test(sess):
    X = cifar.test.images.reshape(10, 1000, 32, 32, 3)
    Y = cifar.test.labels.reshape(10, 1000, 10)
    acc = np.mean([sess.run(accuracy, feed_dict={x: X[i], y_: Y[i], keep_prob: 1.0}) for i in range(10)])
    print("Accuracy: {:.4}%".format(acc * 100))

with tf.Session() as sess:
    
    sess.run(tf.global_variables_initializer())
    
    for i in range(NUM_STEPS):
        batch = cifar.train.next_batch(BATCH_SIZE)
        
        if i % 100 == 0:
            train_accuracy = sess.run(accuracy, feed_dict={x: batch[0], y_: batch[1], keep_prob: 1.0})
            print("step {}, training accuracy {}".format(i, train_accuracy))
            
        sess.run(train_step, feed_dict={x: batch[0], y_: batch[1], keep_prob: 0.5})
        
    test(sess)

step 0, training accuracy 0.1599999964237213
step 100, training accuracy 0.05000000074505806
step 200, training accuracy 0.15000000596046448
step 300, training accuracy 0.15000000596046448
step 400, training accuracy 0.2199999988079071
step 500, training accuracy 0.17000000178813934
step 600, training accuracy 0.25999999046325684
step 700, training accuracy 0.20999999344348907
step 800, training accuracy 0.3199999928474426
step 900, training accuracy 0.38999998569488525
step 1000, training accuracy 0.33000001311302185
step 1100, training accuracy 0.3400000035762787
step 1200, training accuracy 0.3100000023841858
step 1300, training accuracy 0.3799999952316284
step 1400, training accuracy 0.47999998927116394
step 1500, training accuracy 0.33000001311302185
step 1600, training accuracy 0.4300000071525574
step 1700, training accuracy 0.3499999940395355
step 1800, training accuracy 0.4699999988079071
step 1900, training accuracy 0.5400000214576721
step 2000, training accuracy 0.4300000071525574
step 2100, training accuracy 0.4399999976158142
step 2200, training accuracy 0.4300000071525574
step 2300, training accuracy 0.47999998927116394
step 2400, training accuracy 0.5400000214576721
step 2500, training accuracy 0.44999998807907104
step 2600, training accuracy 0.47999998927116394
step 2700, training accuracy 0.4000000059604645
step 2800, training accuracy 0.5
step 2900, training accuracy 0.5600000023841858
step 3000, training accuracy 0.47999998927116394
step 3100, training accuracy 0.4399999976158142
step 3200, training accuracy 0.4099999964237213
step 3300, training accuracy 0.5299999713897705
step 3400, training accuracy 0.5799999833106995
step 3500, training accuracy 0.5099999904632568
step 3600, training accuracy 0.49000000953674316
step 3700, training accuracy 0.5199999809265137
step 3800, training accuracy 0.5400000214576721
step 3900, training accuracy 0.6600000262260437
step 4000, training accuracy 0.5699999928474426
step 4100, training accuracy 0.550000011920929
step 4200, training accuracy 0.46000000834465027
step 4300, training accuracy 0.6000000238418579
step 4400, training accuracy 0.6600000262260437
step 4500, training accuracy 0.550000011920929
step 4600, training accuracy 0.5600000023841858
step 4700, training accuracy 0.5099999904632568
step 4800, training accuracy 0.5899999737739563
step 4900, training accuracy 0.6800000071525574
step 5000, training accuracy 0.5600000023841858
step 5100, training accuracy 0.5400000214576721
step 5200, training accuracy 0.5600000023841858
step 5300, training accuracy 0.6299999952316284
step 5400, training accuracy 0.6899999976158142
step 5500, training accuracy 0.5799999833106995
step 5600, training accuracy 0.5299999713897705
step 5700, training accuracy 0.550000011920929
step 5800, training accuracy 0.6000000238418579
step 5900, training accuracy 0.7099999785423279
step 6000, training accuracy 0.6200000047683716
step 6100, training accuracy 0.6399999856948853
step 6200, training accuracy 0.5099999904632568
step 6300, training accuracy 0.6399999856948853
step 6400, training accuracy 0.7099999785423279
step 6500, training accuracy 0.6499999761581421
step 6600, training accuracy 0.5899999737739563
step 6700, training accuracy 0.5
step 6800, training accuracy 0.6700000166893005
step 6900, training accuracy 0.7300000190734863
step 7000, training accuracy 0.6299999952316284
step 7100, training accuracy 0.6200000047683716
step 7200, training accuracy 0.5400000214576721
step 7300, training accuracy 0.7099999785423279
step 7400, training accuracy 0.7599999904632568
step 7500, training accuracy 0.6499999761581421
step 7600, training accuracy 0.6499999761581421
step 7700, training accuracy 0.5299999713897705
step 7800, training accuracy 0.6499999761581421
step 7900, training accuracy 0.7699999809265137
step 8000, training accuracy 0.6899999976158142
step 8100, training accuracy 0.6200000047683716
step 8200, training accuracy 0.5799999833106995
step 8300, training accuracy 0.6800000071525574
step 8400, training accuracy 0.7699999809265137
step 8500, training accuracy 0.6299999952316284
step 8600, training accuracy 0.6399999856948853
step 8700, training accuracy 0.5199999809265137
step 8800, training accuracy 0.6899999976158142
step 8900, training accuracy 0.699999988079071
step 9000, training accuracy 0.6800000071525574
step 9100, training accuracy 0.6899999976158142
step 9200, training accuracy 0.6200000047683716
step 9300, training accuracy 0.7099999785423279
step 9400, training accuracy 0.7599999904632568
step 9500, training accuracy 0.6000000238418579
step 9600, training accuracy 0.6600000262260437
step 9700, training accuracy 0.550000011920929
step 9800, training accuracy 0.6899999976158142
step 9900, training accuracy 0.7900000214576721
step 10000, training accuracy 0.6000000238418579
step 10100, training accuracy 0.6700000166893005
step 10200, training accuracy 0.5400000214576721
step 10300, training accuracy 0.6600000262260437
step 10400, training accuracy 0.7900000214576721
step 10500, training accuracy 0.6499999761581421
step 10600, training accuracy 0.7099999785423279
step 10700, training accuracy 0.5799999833106995
step 10800, training accuracy 0.6800000071525574
step 10900, training accuracy 0.7900000214576721
step 11000, training accuracy 0.6600000262260437
step 11100, training accuracy 0.6700000166893005
step 11200, training accuracy 0.6000000238418579
step 11300, training accuracy 0.7099999785423279
step 11400, training accuracy 0.7900000214576721
step 11500, training accuracy 0.6800000071525574
step 11600, training accuracy 0.6800000071525574
step 11700, training accuracy 0.550000011920929
step 11800, training accuracy 0.699999988079071
step 11900, training accuracy 0.7599999904632568
step 12000, training accuracy 0.7200000286102295
step 12100, training accuracy 0.6899999976158142
step 12200, training accuracy 0.550000011920929
step 12300, training accuracy 0.7099999785423279
step 12400, training accuracy 0.7699999809265137
step 12500, training accuracy 0.6700000166893005
step 12600, training accuracy 0.699999988079071
step 12700, training accuracy 0.6399999856948853
step 12800, training accuracy 0.7200000286102295
step 12900, training accuracy 0.7599999904632568
step 13000, training accuracy 0.6800000071525574
step 13100, training accuracy 0.7300000190734863
step 13200, training accuracy 0.6399999856948853
step 13300, training accuracy 0.7099999785423279
step 13400, training accuracy 0.7699999809265137
step 13500, training accuracy 0.6700000166893005
step 13600, training accuracy 0.75
step 13700, training accuracy 0.6000000238418579
step 13800, training accuracy 0.6800000071525574
step 13900, training accuracy 0.7900000214576721
step 14000, training accuracy 0.6899999976158142
step 14100, training accuracy 0.6800000071525574
step 14200, training accuracy 0.6200000047683716
step 14300, training accuracy 0.7300000190734863
step 14400, training accuracy 0.8100000023841858
step 14500, training accuracy 0.6600000262260437
step 14600, training accuracy 0.7699999809265137
step 14700, training accuracy 0.6399999856948853
step 14800, training accuracy 0.6899999976158142
step 14900, training accuracy 0.7900000214576721
step 15000, training accuracy 0.6600000262260437
step 15100, training accuracy 0.7300000190734863
step 15200, training accuracy 0.6299999952316284
step 15300, training accuracy 0.7900000214576721
step 15400, training accuracy 0.800000011920929
step 15500, training accuracy 0.6700000166893005
step 15600, training accuracy 0.7300000190734863
step 15700, training accuracy 0.6399999856948853
step 15800, training accuracy 0.7200000286102295
step 15900, training accuracy 0.7599999904632568
step 16000, training accuracy 0.699999988079071
step 16100, training accuracy 0.7799999713897705
step 16200, training accuracy 0.6399999856948853
step 16300, training accuracy 0.7400000095367432
step 16400, training accuracy 0.800000011920929
step 16500, training accuracy 0.6700000166893005
step 16600, training accuracy 0.7699999809265137
step 16700, training accuracy 0.6499999761581421
step 16800, training accuracy 0.7300000190734863
step 16900, training accuracy 0.7699999809265137
step 17000, training accuracy 0.7099999785423279
step 17100, training accuracy 0.7400000095367432
step 17200, training accuracy 0.6600000262260437
step 17300, training accuracy 0.7400000095367432
step 17400, training accuracy 0.7599999904632568
step 17500, training accuracy 0.699999988079071
step 17600, training accuracy 0.7300000190734863
step 17700, training accuracy 0.6600000262260437
step 17800, training accuracy 0.7300000190734863
step 17900, training accuracy 0.7799999713897705
step 18000, training accuracy 0.6700000166893005
step 18100, training accuracy 0.7599999904632568
step 18200, training accuracy 0.7200000286102295
step 18300, training accuracy 0.75
step 18400, training accuracy 0.8100000023841858
step 18500, training accuracy 0.6800000071525574
step 18600, training accuracy 0.7300000190734863
step 18700, training accuracy 0.6700000166893005
step 18800, training accuracy 0.7300000190734863
step 18900, training accuracy 0.7599999904632568
step 19000, training accuracy 0.6899999976158142
step 19100, training accuracy 0.7799999713897705
step 19200, training accuracy 0.6899999976158142
step 19300, training accuracy 0.7200000286102295
step 19400, training accuracy 0.800000011920929
step 19500, training accuracy 0.6800000071525574
step 19600, training accuracy 0.7699999809265137
step 19700, training accuracy 0.699999988079071
step 19800, training accuracy 0.7799999713897705
step 19900, training accuracy 0.7900000214576721
step 20000, training accuracy 0.6800000071525574
step 20100, training accuracy 0.7799999713897705
step 20200, training accuracy 0.75
step 20300, training accuracy 0.7699999809265137
step 20400, training accuracy 0.7699999809265137
step 20500, training accuracy 0.6899999976158142
step 20600, training accuracy 0.7799999713897705
step 20700, training accuracy 0.7300000190734863
step 20800, training accuracy 0.8199999928474426
step 20900, training accuracy 0.7799999713897705
step 21000, training accuracy 0.7599999904632568
step 21100, training accuracy 0.7699999809265137
step 21200, training accuracy 0.75
step 21300, training accuracy 0.8100000023841858
step 21400, training accuracy 0.7900000214576721
step 21500, training accuracy 0.7099999785423279
step 21600, training accuracy 0.7699999809265137
step 21700, training accuracy 0.7200000286102295
step 21800, training accuracy 0.7900000214576721
step 21900, training accuracy 0.800000011920929
step 22000, training accuracy 0.6899999976158142
step 22100, training accuracy 0.7599999904632568
step 22200, training accuracy 0.7099999785423279
step 22300, training accuracy 0.8199999928474426
step 22400, training accuracy 0.7900000214576721
step 22500, training accuracy 0.7699999809265137
step 22600, training accuracy 0.8100000023841858
step 22700, training accuracy 0.7900000214576721
step 22800, training accuracy 0.7900000214576721
step 22900, training accuracy 0.7799999713897705
step 23000, training accuracy 0.75
step 23100, training accuracy 0.75
step 23200, training accuracy 0.7300000190734863
step 23300, training accuracy 0.7900000214576721
step 23400, training accuracy 0.800000011920929
step 23500, training accuracy 0.7300000190734863
step 23600, training accuracy 0.7599999904632568
step 23700, training accuracy 0.7699999809265137
step 23800, training accuracy 0.7400000095367432
step 23900, training accuracy 0.800000011920929
step 24000, training accuracy 0.75
step 24100, training accuracy 0.8100000023841858
step 24200, training accuracy 0.7400000095367432
step 24300, training accuracy 0.7900000214576721
step 24400, training accuracy 0.7900000214576721
step 24500, training accuracy 0.75
step 24600, training accuracy 0.7699999809265137
step 24700, training accuracy 0.699999988079071
step 24800, training accuracy 0.8100000023841858
step 24900, training accuracy 0.8299999833106995
step 25000, training accuracy 0.7099999785423279
step 25100, training accuracy 0.8299999833106995
step 25200, training accuracy 0.7400000095367432
step 25300, training accuracy 0.8100000023841858
step 25400, training accuracy 0.8199999928474426
step 25500, training accuracy 0.7300000190734863
step 25600, training accuracy 0.7699999809265137
step 25700, training accuracy 0.75
step 25800, training accuracy 0.7599999904632568
step 25900, training accuracy 0.8299999833106995
step 26000, training accuracy 0.7400000095367432
step 26100, training accuracy 0.8100000023841858
step 26200, training accuracy 0.7200000286102295
step 26300, training accuracy 0.7400000095367432
step 26400, training accuracy 0.8299999833106995
step 26500, training accuracy 0.7300000190734863
step 26600, training accuracy 0.7699999809265137
step 26700, training accuracy 0.699999988079071
step 26800, training accuracy 0.7599999904632568
step 26900, training accuracy 0.8399999737739563
step 27000, training accuracy 0.7900000214576721
step 27100, training accuracy 0.8199999928474426
step 27200, training accuracy 0.7200000286102295
step 27300, training accuracy 0.7900000214576721
step 27400, training accuracy 0.8399999737739563
step 27500, training accuracy 0.7599999904632568
step 27600, training accuracy 0.75
step 27700, training accuracy 0.7599999904632568
step 27800, training accuracy 0.7900000214576721
step 27900, training accuracy 0.8199999928474426
step 28000, training accuracy 0.7400000095367432
step 28100, training accuracy 0.7799999713897705
step 28200, training accuracy 0.7599999904632568
step 28300, training accuracy 0.7900000214576721
step 28400, training accuracy 0.8100000023841858
step 28500, training accuracy 0.7200000286102295
step 28600, training accuracy 0.7900000214576721
step 28700, training accuracy 0.6899999976158142
step 28800, training accuracy 0.7799999713897705
step 28900, training accuracy 0.800000011920929
step 29000, training accuracy 0.7200000286102295
step 29100, training accuracy 0.800000011920929
step 29200, training accuracy 0.7900000214576721
step 29300, training accuracy 0.7699999809265137
step 29400, training accuracy 0.8399999737739563
step 29500, training accuracy 0.800000011920929
step 29600, training accuracy 0.8199999928474426
step 29700, training accuracy 0.7200000286102295
step 29800, training accuracy 0.7900000214576721
step 29900, training accuracy 0.8100000023841858
step 30000, training accuracy 0.7200000286102295
step 30100, training accuracy 0.8299999833106995
step 30200, training accuracy 0.75
step 30300, training accuracy 0.7900000214576721
step 30400, training accuracy 0.7799999713897705
step 30500, training accuracy 0.7300000190734863
step 30600, training accuracy 0.8100000023841858
step 30700, training accuracy 0.6899999976158142
step 30800, training accuracy 0.7900000214576721
step 30900, training accuracy 0.7900000214576721
step 31000, training accuracy 0.7900000214576721
step 31100, training accuracy 0.8199999928474426
step 31200, training accuracy 0.75
step 31300, training accuracy 0.800000011920929
step 31400, training accuracy 0.8299999833106995
step 31500, training accuracy 0.7799999713897705
step 31600, training accuracy 0.7699999809265137
step 31700, training accuracy 0.75
step 31800, training accuracy 0.7699999809265137
step 31900, training accuracy 0.8100000023841858
step 32000, training accuracy 0.75
step 32100, training accuracy 0.7799999713897705
step 32200, training accuracy 0.7099999785423279
step 32300, training accuracy 0.7900000214576721
step 32400, training accuracy 0.7900000214576721
step 32500, training accuracy 0.7300000190734863
step 32600, training accuracy 0.7900000214576721
step 32700, training accuracy 0.7400000095367432
step 32800, training accuracy 0.75
step 32900, training accuracy 0.8199999928474426
step 33000, training accuracy 0.75
step 33100, training accuracy 0.8600000143051147
step 33200, training accuracy 0.7699999809265137
step 33300, training accuracy 0.800000011920929
step 33400, training accuracy 0.8299999833106995
step 33500, training accuracy 0.7699999809265137
step 33600, training accuracy 0.8100000023841858
step 33700, training accuracy 0.75
step 33800, training accuracy 0.7799999713897705
step 33900, training accuracy 0.7799999713897705
step 34000, training accuracy 0.7599999904632568
step 34100, training accuracy 0.7699999809265137
step 34200, training accuracy 0.75
step 34300, training accuracy 0.75
step 34400, training accuracy 0.8600000143051147
step 34500, training accuracy 0.7400000095367432
step 34600, training accuracy 0.800000011920929
step 34700, training accuracy 0.699999988079071
step 34800, training accuracy 0.800000011920929
step 34900, training accuracy 0.7799999713897705
step 35000, training accuracy 0.8100000023841858
step 35100, training accuracy 0.800000011920929
step 35200, training accuracy 0.75
step 35300, training accuracy 0.7699999809265137
step 35400, training accuracy 0.7900000214576721
step 35500, training accuracy 0.7799999713897705
step 35600, training accuracy 0.8199999928474426
step 35700, training accuracy 0.7699999809265137
step 35800, training accuracy 0.7699999809265137
step 35900, training accuracy 0.7799999713897705
step 36000, training accuracy 0.7400000095367432
step 36100, training accuracy 0.7599999904632568
step 36200, training accuracy 0.75
step 36300, training accuracy 0.7799999713897705
step 36400, training accuracy 0.8100000023841858
step 36500, training accuracy 0.7400000095367432
step 36600, training accuracy 0.8199999928474426
step 36700, training accuracy 0.7400000095367432
step 36800, training accuracy 0.7699999809265137
step 36900, training accuracy 0.7900000214576721
step 37000, training accuracy 0.7699999809265137
step 37100, training accuracy 0.7699999809265137
step 37200, training accuracy 0.75
step 37300, training accuracy 0.800000011920929
step 37400, training accuracy 0.7900000214576721
step 37500, training accuracy 0.7599999904632568
step 37600, training accuracy 0.8199999928474426
step 37700, training accuracy 0.7599999904632568
step 37800, training accuracy 0.7599999904632568
step 37900, training accuracy 0.7900000214576721
step 38000, training accuracy 0.7900000214576721
step 38100, training accuracy 0.7799999713897705
step 38200, training accuracy 0.75
step 38300, training accuracy 0.8100000023841858
step 38400, training accuracy 0.800000011920929
step 38500, training accuracy 0.75
step 38600, training accuracy 0.7900000214576721
step 38700, training accuracy 0.7400000095367432
step 38800, training accuracy 0.7799999713897705
step 38900, training accuracy 0.7699999809265137
step 39000, training accuracy 0.7400000095367432
step 39100, training accuracy 0.8500000238418579
step 39200, training accuracy 0.7099999785423279
step 39300, training accuracy 0.7900000214576721
step 39400, training accuracy 0.8199999928474426
step 39500, training accuracy 0.8199999928474426
step 39600, training accuracy 0.7900000214576721
step 39700, training accuracy 0.7599999904632568
step 39800, training accuracy 0.7699999809265137
step 39900, training accuracy 0.8100000023841858
step 40000, training accuracy 0.8299999833106995
step 40100, training accuracy 0.8199999928474426
step 40200, training accuracy 0.7400000095367432
step 40300, training accuracy 0.7400000095367432
step 40400, training accuracy 0.800000011920929
step 40500, training accuracy 0.8299999833106995
step 40600, training accuracy 0.8600000143051147
step 40700, training accuracy 0.7200000286102295
step 40800, training accuracy 0.7599999904632568
step 40900, training accuracy 0.800000011920929
step 41000, training accuracy 0.7900000214576721
step 41100, training accuracy 0.800000011920929
step 41200, training accuracy 0.7400000095367432
step 41300, training accuracy 0.8100000023841858
step 41400, training accuracy 0.7900000214576721
step 41500, training accuracy 0.7900000214576721
step 41600, training accuracy 0.7900000214576721
step 41700, training accuracy 0.7900000214576721
step 41800, training accuracy 0.8100000023841858
step 41900, training accuracy 0.7900000214576721
step 42000, training accuracy 0.7900000214576721
step 42100, training accuracy 0.8100000023841858
step 42200, training accuracy 0.7799999713897705
step 42300, training accuracy 0.8299999833106995
step 42400, training accuracy 0.8199999928474426
step 42500, training accuracy 0.7799999713897705
step 42600, training accuracy 0.8100000023841858
step 42700, training accuracy 0.7799999713897705
step 42800, training accuracy 0.7799999713897705
step 42900, training accuracy 0.8199999928474426
step 43000, training accuracy 0.7799999713897705
step 43100, training accuracy 0.8700000047683716
step 43200, training accuracy 0.7799999713897705
step 43300, training accuracy 0.8100000023841858
step 43400, training accuracy 0.7799999713897705
step 43500, training accuracy 0.8299999833106995
step 43600, training accuracy 0.8299999833106995
step 43700, training accuracy 0.800000011920929
step 43800, training accuracy 0.7799999713897705
step 43900, training accuracy 0.800000011920929
step 44000, training accuracy 0.800000011920929
step 44100, training accuracy 0.8100000023841858
step 44200, training accuracy 0.7900000214576721
step 44300, training accuracy 0.75
step 44400, training accuracy 0.8199999928474426
step 44500, training accuracy 0.7799999713897705
step 44600, training accuracy 0.800000011920929
step 44700, training accuracy 0.8100000023841858
step 44800, training accuracy 0.7799999713897705
step 44900, training accuracy 0.7599999904632568
step 45000, training accuracy 0.800000011920929
step 45100, training accuracy 0.7799999713897705
step 45200, training accuracy 0.7799999713897705
step 45300, training accuracy 0.800000011920929
step 45400, training accuracy 0.8100000023841858
step 45500, training accuracy 0.8100000023841858
step 45600, training accuracy 0.8199999928474426
step 45700, training accuracy 0.7900000214576721
step 45800, training accuracy 0.8199999928474426
step 45900, training accuracy 0.8199999928474426
step 46000, training accuracy 0.7900000214576721
step 46100, training accuracy 0.8199999928474426
step 46200, training accuracy 0.7599999904632568
step 46300, training accuracy 0.8100000023841858
step 46400, training accuracy 0.8100000023841858
step 46500, training accuracy 0.800000011920929
step 46600, training accuracy 0.800000011920929
step 46700, training accuracy 0.7799999713897705
step 46800, training accuracy 0.7900000214576721
step 46900, training accuracy 0.7799999713897705
step 47000, training accuracy 0.7099999785423279
step 47100, training accuracy 0.8500000238418579
step 47200, training accuracy 0.7699999809265137
step 47300, training accuracy 0.7699999809265137
step 47400, training accuracy 0.800000011920929
step 47500, training accuracy 0.8199999928474426
step 47600, training accuracy 0.8199999928474426
step 47700, training accuracy 0.800000011920929
step 47800, training accuracy 0.800000011920929
step 47900, training accuracy 0.7599999904632568
step 48000, training accuracy 0.800000011920929
step 48100, training accuracy 0.800000011920929
step 48200, training accuracy 0.8199999928474426
step 48300, training accuracy 0.8299999833106995
step 48400, training accuracy 0.8100000023841858
step 48500, training accuracy 0.8100000023841858
step 48600, training accuracy 0.7900000214576721
step 48700, training accuracy 0.7699999809265137
step 48800, training accuracy 0.75
step 48900, training accuracy 0.8199999928474426
step 49000, training accuracy 0.8199999928474426
step 49100, training accuracy 0.7599999904632568
step 49200, training accuracy 0.8199999928474426
step 49300, training accuracy 0.800000011920929
step 49400, training accuracy 0.8199999928474426
step 49500, training accuracy 0.7799999713897705
step 49600, training accuracy 0.8199999928474426
step 49700, training accuracy 0.8199999928474426
step 49800, training accuracy 0.7900000214576721
step 49900, training accuracy 0.8100000023841858
Accuracy: 76.33%

本文内容由网友自发贡献，版权归原作者所有，本站不承担相应法律责任。如您发现有涉嫌抄袭侵权的内容，请联系:hwhale#tublm.com(使用前将#替换为@)

卷积神经网络处理Cifar-10分类数据集的相关文章

cmake在Windows上的使用

cmake介绍 cmake是一个构建C 43 43 代码的跨平台工具 xff0c 他能够干什么呢 xff1f 他能够搜索你的平台环境 xff0c 然后生成平台上用于Build的文件在Windows上安装了Visual Studio 他能够
ubuntu系统安装完nvidia显卡驱动后黑屏，不能进入系统

昨天想看显卡 xff0c 更新了下驱动 xff0c 发现服务器重启进不去步骤 1 开机按esc 进入选项界面2 进去以后选择一个括号里面带recovery mode的选项3 然后它自动黑屏出现代码 xff0c 然后弹出一个选择框 xff
ubuntu18.04下cmake的安装

一使用安装命令 span class token function sudo span span class token function apt span span class token function install span c
Nginx/Lighttpd/Apache三大web服务器对比分析

一 xff0e 软件介绍 apache lighttpd nginx 1 lighttpd Lighttpd是一个具有非常低的内存开销 xff0c cpu占用率低 xff0c 效能好 xff0c 以及丰富的模块等特点 lighttpd是众多
STM32学习笔记八——串口实现不定长数据传输

生命不息 xff0c 奋斗不止 0 摘要今天说一下STM32单片机的接收不定长度字节数据的方法由于STM32单片机带IDLE中断 xff0c 所以利用这个中断 xff0c 可以接收不定长字节的数据 xff0c 由于STM32属于ARM单
linux shell 字符串长度

本篇内容主要包括以下内容 xff1a 获取字符串长度的方法各方法间性能比较相关内容补充说明 xff1a 本篇内容为学习shell 编程老男孩教程笔记 xff0c 部分内容为相关视频课程内容 xff0c 补充部分是我查资料所得 xf
measurement studio测控与单片机通信上位机

using NationalInstruments Analysis using NationalInstruments Analysis Conversion using NationalInstruments Analysis Dsp
C语言实现Socket简单通信

环境是linux xff0c 不过应该没什么影响 xff0c 因为只用到了socket的基本用法 xff0c 没有涉及pthread等分为服务器端和客户端 xff0c 服务器端监听端口发来的请求 xff0c 收到后向客户端发送一个Hell
接口工具POSTMAN怎么设置中文

相信各位测试工程师或者测试小白们都会在进行接口测试的时候接触到POSTMAN这款强大的接口测试软件 Postman 从最初设计上就是为接口测试而生的 xff0c 所以在程序员中广泛使用 xff0c 在开发调试网络程序时跟踪一些网络请求 xf
【第三方库】libcurl实现http及https方式下载和访问

准备条件 xff1a 编译好的libcurl库 xff0c 如果要支持https xff0c 需要和openssl一起编译 xff0c 网上教程较多示例代码 xff1a xff08 使用loadlibrary的方式只是为了测试方便 xff
ROS学习系列：第四讲 — gazebo模型建立

gazebo模型元素 World Files 文件包含机器人灯光传感器静态物体文件格式SDF xff0c 后缀扩展名 world gazebo服务端gzserver读取world文件产生和构成world Model Files 文件
2021电赛F题视觉教程+代码免费开源

2021电赛F题视觉教程 43 代码免费开源最近好多要电赛题的源码 xff0c 其他csdn营销号下载都需要会员或钱 xff0c 正好最近课设又要做一遍电赛小车题 xff0c 哥们先把代码开源了 xff0c 饿死营销号电赛宝藏链接 xf
2021全国大学生电子设计竞赛论文（智能送药小车（F题））（电赛论文模板）

2021全国大学生电子设计竞赛论文 xff08 智能送药小车 xff08 F题 xff09 xff09 写在前面的话电赛是一个很奇妙的过程 xff0c 可能有些人觉得电赛的门槛太高 xff0c 那便意味着 xff0c 当你决定要参加电赛的
gazebo9中在墙上添加二维码

1 aruco 1到aruco 34这些二维码图片 xff08 png格式 xff09 及下述配置文件上传至https download csdn net download zhaohaowu 83982724 将这34张二维码图片保存到
最新MAC从官网安装QT Creator安装以及QT环境的配置(超详细)

先开喷 xff0c 垃圾QT xff0c 官网的下载什么玩意儿 xff0c 搞了半天都弄不好 xff0c 后来琢磨了很久才装成功首先直接访问官网的下载网站 xff1a https www qt io download qt install
ROS下如何将GPS数据在卫星地图显示（两种开源方法）

自己知道的两种方法 xff0c 分别基于rviz satellite和mapviz xff0c 实测下来Mapviz在清晰度和准确度上效果感觉更好一些 xff0c 大家可以都试下 xff0c 有更好的方法可以留言大家讨论下一 rviz s
JAVA类加载器包括几种？它们之间的关系是怎么样的？双亲委派机制是什么意思？有什么好处？

JAVA类加载器包括几种 xff1f 引导类加载器 bootstrap class loader 启动类加载器主要加载的是JVM自身需要的类 xff0c 这个类加载使用C 43 43 语言实现的 xff0c 是虚拟机自身的一部分 xff0c
微信小程序登录

login wxml lt view class 61 34 content 34 gt lt 用户名 gt lt view class 61 34 username 34 gt lt view class 61 34 title 34 g
小程序-用户注册

wxml lt form bindsubmit 61 34 formSubmit 34 gt lt view class 61 34 user 34 gt lt view class 61 34 hr 34 gt lt view gt lt
小程序-滑动导航

home1 wxml lt pages home1 home1 wxml gt lt 滑动导航 gt lt view class 61 34 bg 34 gt lt view class 61 34 nav 34 gt lt scroll

随机推荐

小程序-地图（腾讯地图）

map wxml 经度 xff1b 维度 xff1b 缩放比例 xff1b 样式 xff1b 显示位置 lt pages map map wxml gt lt 地图组件 gt lt 经度 xff1b 维度 xff1b 缩放比例 xff1b
nginx和php之间是怎样通信的呢

FastCGI原理 FastCGI是一个运用于Http Server和动态脚本语言间通信的接口 xff0c 多数流行的Http Server都支持FastCGI xff0c 包括Apache Nginx和lighttpd等同时 xff0c
MYSQL(分表)千万级数据量的优化方法积累

1 分库分表一个主表 xff08 例如用户表 xff09 无限制的增长势必严重影响性能 xff0c 分库与分表是一个很不错的解决途径 xff0c 也就是性能优化途径 xff0c 现在的案例是我们有一个1000多万条记录的用户表member
GO 字符串转换为 JSON数据与各种转换

span class hljs keyword style color rgb 51 51 51 font weight bold https segmentfault com q 1010000005025933 a 1020000005
5种Redis数据结构详解

http www php cn php weizijiaocheng 388126 html 本文我们主要和大家分享 5种Redis数据结构详解 xff0c 希望文中的案例和代码 xff0c 能帮助到大家 2 1 1 全局命令 1 查看所有
Ubuntu解决Vscode检测到#include错误，请更新includepath

第一步终端输入gcc v E x c 找到路径 usr lib gcc x86 64 linux gnu 11 include usr local include usr include x86 64 linux gnu usr incl
Redis内部数据结构详解(4)——ziplist

本文是 Redis内部数据结构详解系列的第四篇在本文中 xff0c 我们首先介绍一个新的Redis内部数据结构 ziplist xff0c 然后在文章后半部分我们会讨论一下在robj dict和ziplist的基础上 xff0c Red
golang是如何实现高并发的？深入领会MPG模式

多线程共享内存 xff0c 这也是Java C 或者C 43 43 等语言中的多线程开发的常规方法 xff0c 其实golang语言也支持这种传统模式 xff0c 另外一种是Go语言特有的 xff0c 也是Go语言推荐的 xff1a CSP
Mysql索引类型Btree和Hash的区别以及使用场景

遇到单表数据量大的时候很多开发者都会想到给相对的字段建立索引来提高性能 xff08 mysql索引的使用 xff09 xff0c 但很少会去关注索引的类型该如何选择 xff0c 在mysql中支持有两种类型 xff0c 最常用的也是默认的B
tcpreplay的安装使用

转自 xff1a https www cnblogs com zlslch p 7325599 html utm source 61 itdadao amp utm medium 61 referral tcpreplay是什么 xff1f
Ubuntu18.04从零配置到zed2i实现，orb-slam3运行，ros安装（Ubuntu18.04 3050ti）系列一：cuda与csdnn安装。

利用双系统来安装Ubuntu18 04 采用的是U盘烧录镜像 xff0c 硬盘为980 256G 目录 1 烧录镜像以及分区 2 设置ubuntu密码 3 网络认证 4 更新显卡驱动以及软件源 5 安装搜狗输入法 6 安装vpn 7 安装C
java ee 话外之 http

HTTP请求格式当浏览器向web服务器发出请求时 xff0c 它向服务器传递了一个数据块 xff0c 也就是请求信息 xff0c http请求信息由三个部分组成 1 请求方法 url协议版本 2 请求头 xff08 request he
Pytorch学习(3) —— nn.Parameter nn.ParameterList nn.ParameterDict 源码解析

为了更好理解Pytorch基本类的实现方法 xff0c 我这里给出了关于参数方面的3个类的源码详解此部分可以更好的了解实现逻辑结构 xff0c 有助于后续代码理解 xff0c 学pytorch的话这个不是必须掌握的 xff0c 看不懂也没
针对电陶炉E5错误的维修总结（狗头）

一个编程技术员开始研究电陶炉维修是不是有些奇怪没办法 xff0c 最近家里面各种电器都开始坏掉了 xff0c 有的是按钮 xff0c 有的是断线 xff0c 有的就是电路板内部故障固件坏了买相应零件修好就能用 xff0c 比如用4个开关
关于视觉SLAM十四讲sophus库安装报错

Sophus安装 xff1a git clone https github com strasdat sophus git cd sophus mkdir build cmake make 这时候系统报错 error lvalue requ
[BPU部署教程] 教你搞定YOLOV5部署 (版本: 6.2)

最近一些人问我怎么在BPU上部署yolov5 xff0c 因为之前的博客 BPU部署教程一文带你轻松走出模型部署新手村介绍的网络都是基于Caffe的 xff0c 自己的网络都是基于pytorch的 xff0c 所以遇到了很多坑鉴于这些需
在旭日X3派开发板上使用USB Wifi来提高网络速度

对于我来说 xff0c 开发板自带的wifi模块速度不是很满意 xff0c 下载或者传文件啥的最多也就2M s 而且 xff0c 在之前测评博客首发多方位玩转地平线新发布AIoT开发板旭日X3派 Sunrise x3 Pi 插电 x
体验极速——在旭日X3派上使用双频1300M USB无线网卡

上一篇博客在旭日X3派开发板上使用USB Wifi来提高网络速度提供一种低成本 xffe5 20的USB Wifi解决方案这个模块的传输速度在10M s以内 xff0c 尽管满足正常的开发需求 xff0c 但在项目应用时 xff0c
linux深度学习服务器搭建——CUDA与cuDNN的选择与安装

前言本文章参考实验室师妹的文章Ubuntu14 04 43 CUDA8 0 43 Opencv3 1 43 Anaconda2 43 Caffe安装 xff0c 最近安装最新版时候遇到不少坑 xff0c 下面就介绍下如何去安装CUDA和c
卷积神经网络处理Cifar-10分类数据集

Cifar 10分类数据集 Cifar 10分类数据集简介 CIFAR 10数据集由10个类的60000个32x32彩色图像组成 xff0c 每个类有6000个图像有50000个训练图像和10000个测试图像数据集分为五个训练批次和一个

卷积神经网络处理Cifar-10分类数据集

Cifar-10分类数据集

Cifar-10分类数据集简介

代码

1. 定义Cifar-10类

2. 定义函数

3. 定义数据管理器类

4. 显示Cifar-10数据集图片

4.1 Cifar-10数据集显示

5. 神经网络

5.1 定义函数

5.2 网络结构

5.2.1 双层卷积网络

5.2.2 三层卷积网络

5.2.3 多层卷积网络

5.3 训练设定

5.4 会话

卷积神经网络处理Cifar-10分类数据集 的相关文章

随机推荐

热门标签

卷积神经网络处理Cifar-10分类数据集的相关文章