用python将数据集图像分为10类，每类的训练集和测试集数量相同，图片大小为32*32

以下是一个简单的代码示例，使用Keras库将CIFAR-10数据集（包含10类32x32像素图像）分为训练集和测试集，并进行数据预处理：

import numpy as np
from keras.datasets import cifar10
from keras.utils import to_categorical

# load the CIFAR-10 dataset
(x_train, y_train), (x_test, y_test) = cifar10.load_data()

# convert labels to one-hot encoding
y_train = to_categorical(y_train, num_classes=10)
y_test = to_categorical(y_test, num_classes=10)

# preprocess data by scaling pixel values to [0, 1] and converting to float32
x_train = x_train.astype('float32') / 255
x_test = x_test.astype('float32') / 255

# split data into 10 classes
class_indices = [np.where(y_train[:, i] == 1)[0] for i in range(10)]
train_indices = []
test_indices = []
for indices in class_indices:
    np.random.shuffle(indices)
    split_point = int(0.8 * len(indices))
    train_indices += indices[:split_point].tolist()
    test_indices += indices[split_point:].tolist()

# create training and testing datasets
x_train_new = x_train[train_indices]
y_train_new = y_train[train_indices]
x_test_new = x_test[test_indices]
y_test_new = y_test[test_indices]

在这个示例中，我们首先加载CIFAR-10数据集，并将标签转换为one-hot编码。然后，我们对图像数据进行预处理，将像素值缩放到[0,1]范围内，并将数据类型转换为float32。接下来，我们根据类别将数据集分为10个子集，并将每个子集的80％用作训练数据，20％用作测试数据。最后，我们创建新的训练和测试数据集，只包含我们选择的子集。