将自己C盘中不规则图片作为深度学习模型输入，其中包括训练集、测试集、校验集，图片个数分别为7000、2000、1000.实现将动态卷积引入到densenet-inception代码实现

首先，将不规则图片按照一定的方式划分为训练集、测试集和验证集，并将它们存储在不同的文件夹中。然后，我们需要使用Python中的Keras库来实现densenet-inception模型，并添加动态卷积层。

以下是示例代码：

from keras.layers import Input, Dense, Conv2D, MaxPooling2D, GlobalAveragePooling2D, Dropout, Concatenate, BatchNormalization, Activation
from keras.models import Model
from keras.optimizers import Adam
from keras.regularizers import l2
import os

# 定义输入形状
inputs = Input(shape=(224, 224, 3))

# 定义densenet-inception模型
def densenet_inception(inputs):
    # convolution block 1
    x = Conv2D(64, (3,3), padding='same', kernel_regularizer=l2(0.01))(inputs)
    x = BatchNormalization()(x)
    x = Activation('relu')(x)

    # dense block 1
    c1 = Concatenate()([inputs, x])
    x = Conv2D(128, (1,1), padding='same', kernel_regularizer=l2(0.01))(c1)
    x = BatchNormalization()(x)
    x = Activation('relu')(x)
    x = Conv2D(32, (3,3), padding='same', kernel_regularizer=l2(0.01))(x)
    x = BatchNormalization()(x)
    x = Activation('relu')(x)

    # inception block 1
    c2 = Concatenate()([inputs, x])
    i1_1 = Conv2D(64, (1,1), padding='same', kernel_regularizer=l2(0.01))(c2)
    i1_1 = BatchNormalization()(i1_1)
    i1_1 = Activation('relu')(i1_1)
    i1_3 = Conv2D(64, (1,1), padding='same', kernel_regularizer=l2(0.01))(c2)
    i1_3 = BatchNormalization()(i1_3)
    i1_3 = Activation('relu')(i1_3)
    i1_3 = Conv2D(64, (3,3), padding='same', kernel_regularizer=l2(0.01))(i1_3)
    i1_3 = BatchNormalization()(i1_3)
    i1_3 = Activation('relu')(i1_3)
    i1_5 = Conv2D(64, (1,1), padding='same', kernel_regularizer=l2(0.01))(c2)
    i1_5 = BatchNormalization()(i1_5)
    i1_5 = Activation('relu')(i1_5)
    i1_5 = Conv2D(64, (5,5), padding='same', kernel_regularizer=l2(0.01))(i1_5)
    i1_5 = BatchNormalization()(i1_5)
    i1_5 = Activation('relu')(i1_5)
    i1_7 = MaxPooling2D((3,3), strides=(1,1), padding='same')(c2)
    i1_7 = Conv2D(64, (1,1), padding='same', kernel_regularizer=l2(0.01))(i1_7)
    i1_7 = BatchNormalization()(i1_7)
    i1_7 = Activation('relu')(i1_7)
    x = Concatenate()([i1_1, i1_3, i1_5, i1_7])

    # dynamic convolution block 1
    d1 = Conv2D(256, (1,1), padding='same', kernel_regularizer=l2(0.01))(x)
    d1 = BatchNormalization()(d1)
    d1 = Activation('relu')(d1)
    d1 = Conv2D(256, (3,3), padding='same', kernel_regularizer=l2(0.01))(d1)
    d1 = BatchNormalization()(d1)
    d1 = Activation('relu')(d1)
    d1 = Conv2D(1024, (1,1), padding='same', kernel_regularizer=l2(0.01))(d1)
    d1 = BatchNormalization()(d1)
    d1 = Activation('relu')(d1)
    d1 = Dropout(0.5)(d1)

    # dense block 2
    c3 = Concatenate()([x, d1])
    x = Conv2D(256, (1,1), padding='same', kernel_regularizer=l2(0.01))(c3)
    x = BatchNormalization()(x)
    x = Activation('relu')(x)
    x = Conv2D(32, (3,3), padding='same', kernel_regularizer=l2(0.01))(x)
    x = BatchNormalization()(x)
    x = Activation('relu')(x)

    # inception block 2
    c4 = Concatenate()([x, d1])
    i2_1 = Conv2D(64, (1,1), padding='same', kernel_regularizer=l2(0.01))(c4)
    i2_1 = BatchNormalization()(i2_1)
    i2_1 = Activation('relu')(i2_1)
    i2_3 = Conv2D(64, (1,1), padding='same', kernel_regularizer=l2(0.01))(c4)
    i2_3 = BatchNormalization()(i2_3)
    i2_3 = Activation('relu')(i2_3)
    i2_3 = Conv2D(64, (3,3), padding='same', kernel_regularizer=l2(0.01))(i2_3)
    i2_3 = BatchNormalization()(i2_3)
    i2_3 = Activation('relu')(i2_3)
    i2_5 = Conv2D(64, (1,1), padding='same', kernel_regularizer=l2(0.01))(c4)
    i2_5 = BatchNormalization()(i2_5)
    i2_5 = Activation('relu')(i2_5)
    i2_5 = Conv2D(64, (5,5), padding='same', kernel_regularizer=l2(0.01))(i2_5)
    i2_5 = BatchNormalization()(i2_5)
    i2_5 = Activation('relu')(i2_5)
    i2_7 = MaxPooling2D((3,3), strides=(1,1), padding='same')(c4)
    i2_7 = Conv2D(64, (1,1), padding='same', kernel_regularizer=l2(0.01))(i2_7)
    i2_7 = BatchNormalization()(i2_7)
    i2_7 = Activation('relu')(i2_7)
    x = Concatenate()([i2_1, i2_3, i2_5, i2_7])

    # dynamic convolution block 2
    d2 = Conv2D(256, (1,1), padding='same', kernel_regularizer=l2(0.01))(x)
    d2 = BatchNormalization()(d2)
    d2 = Activation('relu')(d2)
    d2 = Conv2D(256, (3,3), padding='same', kernel_regularizer=l2(0.01))(d2)
    d2 = BatchNormalization()(d2)
    d2 = Activation('relu')(d2)
    d2 = Conv2D(1024, (1,1), padding='same', kernel_regularizer=l2(0.01))(d2)
    d2 = BatchNormalization()(d2)
    d2 = Activation('relu')(d2)
    d2 = Dropout(0.5)(d2)

    # global average pooling and output layer
    x = Concatenate()([x, d2])
    x = GlobalAveragePooling2D()(x)
    x = Dense(10, activation='softmax')(x)

    return x

# 构建模型
outputs = densenet_inception(inputs)
model = Model(inputs=inputs, outputs=outputs)

# 编译模型
model.compile(optimizer=Adam(lr=0.001), loss='categorical_crossentropy', metrics=['accuracy'])

# 加载数据
train_dir = 'path/to/train/dir'
test_dir = 'path/to/test/dir'
val_dir = 'path/to/val/dir'

# 图像生成器
from keras.preprocessing.image import ImageDataGenerator

train_datagen = ImageDataGenerator(rescale=1./255, shear_range=0.2, zoom_range=0.2, horizontal_flip=True)
test_datagen = ImageDataGenerator(rescale=1./255)
val_datagen = ImageDataGenerator(rescale=1./255)

train_generator = train_datagen.flow_from_directory(train_dir, target_size=(224,224), batch_size=32, class_mode='categorical')
test_generator = test_datagen.flow_from_directory(test_dir, target_size=(224,224), batch_size=32, class_mode='categorical')
val_generator = val_datagen.flow_from_directory(val_dir, target_size=(224,224), batch_size=32, class_mode='categorical')

# 训练模型
model.fit_generator(train_generator, steps_per_epoch=7000//32, epochs=10, validation_data=val_generator, validation_steps=1000//32)

# 评估模型
score = model.evaluate_generator(test_generator, steps=2000//32)
print("Test accuracy:", score[1])

在上面的代码中，我们定义了一个densenet-inception模型，它包括了动态卷积层。我们还使用Keras的ImageDataGenerator类来加载和增强数据。最后，我们使用fit_generator()函数来训练模型，并使用evaluate_generator()函数来评估模型的性能。

请注意，这只是一种示例实现，您可能需要根据您的具体需求进行修改。