2023年安徽省大数据与人工智能应用竞赛

人工智能赛道（网络赛-本科组）-答题卡

2023年9月

注：请将每一道题的所写源代码与运行结果截图保存到答题卡对应位置；

将答题卡以队伍编码.docx形式命名在系统进行提交。

第一部分：人工智能基础环境搭建部署（10分）

注：任务1与任务2任选一题完成即可。

o任务1

o任务2

第二部分：样本数据预处理（30分）

注：任务1与任务2都需要完成。

o任务1（15分）。

import pandas as pd

data = pd.read_csv("task2_1.csv")

data = data.dropna(subset=['WORK_PROVINCE'])

non_alphanumeric_count = data[data['WORK_PROVINCE'].str.isalnum()]['WORK_PROVINCE'].count()

print("非字母组合代码的样本个数:", non_alphanumeric_count)

top_20_avg = data.nlargest(20, 'AVG_FLIGHT_COUNT')['AVG_FLIGHT_COUNT'].mean()

print("AVG_FLIGHT_COUNT列最大的20个数值的平均值:", top_20_avg)

o任务2（15分）。

import cv2  

import numpy as np  

img = cv2.imread('data/task2/task2_2.jpg')  

img = cv2.resize(img, (650, 360))  

img1 = img[:, 325:]  

img2 = cv2.cvtColor(img1, cv2.COLOR_BGR2YCrCb)  

img3 = np.zeros_like(img2)  

for i in range(img2.shape[0]):  

    for j in range(img2.shape[1]):  

        if 133 <= img2[i, j][1] <= 173 and 77 <= img2[i, j][2] <= 127:  

            img3[i, j] = [255, 255, 255]  

cv2.imshow('img1', img1)  

cv2.imshow('img3', img3)  

cv2.waitKey(0)  

cv2.destroyAllWindows()

第三部分：传统机器学习算法设计及应用（20分）

import os

import cv2

import numpy as np

from sklearn.cluster import KMeans

from sklearn.svm import SVC

from sklearn.model_selection import train_test_split

from sklearn.metrics import precision_score, recall_score, f1_score

from sklearn.preprocessing import StandardScaler

def extract_sift_features(images, num_clusters):

    sift = cv2.SIFT_create()

    descriptors = []

    for image in images:

        kp, desc = sift.detectAndCompute(image, None)

        if desc is not None:

            descriptors.extend(desc)

    descriptors = np.array(descriptors)

    kmeans = KMeans(n_clusters=num_clusters)

    kmeans.fit(descriptors)

    return kmeans

data_dir = "data/task3"

categories = ["cat", "dog"]

num_clusters = 100  

images = []

labels = []

for category_idx, category in enumerate(categories):

    category_dir = os.path.join(data_dir, category)

    for filename in os.listdir(category_dir):

        if filename.endswith(".jpg"):

            image_path = os.path.join(category_dir, filename)

            image = cv2.imread(image_path, cv2.IMREAD_GRAYSCALE)

            images.append(image)

            labels.append(category_idx)

images = np.array(images)

labels = np.array(labels)

kmeans = extract_sift_features(images, num_clusters)

def image_to_bow(image, kmeans_model):

    kp, desc = sift.detectAndCompute(image, None)

    if desc is not None:

        bow = kmeans_model.predict(desc)

        bow_hist = np.bincount(bow, minlength=num_clusters)

        return bow_hist

    else:

        return np.zeros(num_clusters)

X = []

for image in images:

    bow_vector = image_to_bow(image, kmeans)

    X.append(bow_vector)

X = np.array(X)

X_train, X_test, y_train, y_test = train_test_split(X, labels, test_size=0.1, random_state=42)

scaler = StandardScaler()

X_train = scaler.fit_transform(X_train)

X_test = scaler.transform(X_test)

svm_model = SVC(kernel='linear')

svm_model.fit(X_train, y_train)

y_pred = svm_model.predict(X_test)

precision = precision_score(y_test, y_pred, average='weighted')

recall = recall_score(y_test, y_pred, average='weighted')

f1 = f1_score(y_test, y_pred, average='weighted')

print(f"Precision: {precision:.2f}")

print(f"Recall: {recall:.2f}")

print(f"F1-score: {f1:.2f}")

第四部分：深度学习算法设计及应用（20分）

import numpy as np  

from tensorflow.keras.models import Sequential  

from tensorflow.keras.layers import Dense, Conv2D, MaxPooling2D, Flatten  

from tensorflow.keras.preprocessing.image import ImageDataGenerator  

from tensorflow.keras.optimizers import SGD

import cv2

# 读取训练集和测试集的路径以及标签  

train_path = np.loadtxt('trainlabels.txt', delimiter=' ',encoding='gbk',dtype='str')  

test_path = np.loadtxt('testlabels.txt', delimiter=' ',encoding='gbk',dtype='str')  

train_labels = train_path[:, 1:]  

train_path = train_path[:, 0]  

test_labels = test_path[:, 1:]  

test_path = test_path[:, 0]  

def preprocess_data(img_path, labels):  

    img = cv2.imread(img_path)  

    img = cv2.resize(img,(64, 64))   

    img = np.array(img)  

    img = img / 255.0  # 归一化  

    return img, labels  

train_data, train_labels = zip(*[preprocess_data(train_path[i], train_labels[i]) for i in range(len(train_path))])  

test_data, test_labels = zip(*[preprocess_data(test_path[i], test_labels[i]) for i in range(len(test_path))])  

train_data = np.array(train_data)  

train_labels = np.array(train_labels,dtype='int')  

test_data = np.array(test_data)  

test_labels = np.array(test_labels,dtype='int')  

train_data = train_data.reshape((train_data.shape[0], 64, 64, 3))  

test_data = test_data.reshape((test_data.shape[0], 64, 64, 3))  

model = Sequential()  

model.add(Conv2D(32, kernel_size=(3, 3), activation='relu', input_shape=(64, 64, 3)))  

model.add(MaxPooling2D(pool_size=(2, 2)))  

model.add(Conv2D(64, (3, 3), activation='relu'))  

model.add(MaxPooling2D(pool_size=(2, 2)))  

model.add(Flatten())  

model.add(Dense(128, activation='relu'))  

model.add(Dense(7, activation='sigmoid'))  

model.compile(loss='mean_squared_error', optimizer=SGD(learning_rate=0.001,decay=0.0001), metrics=['accuracy'])  

model.fit(train_data, train_labels, epochs=100, batch_size=32)  

loss, accuracy = model.evaluate(test_data, test_labels)  

print('Test accuracy: %.2f' % (accuracy*100))

第五部分：人工智能技术综合应用（20分）

import jieba

import jieba

from jieba.analyse import extract_tags

import codecs

with codecs.open('task5data.txt', 'r', encoding='utf-8') as file:

    text = file.read()

keywords = extract_tags(text, topK=30, withWeight=True)

keyword_dict = {key: weight for key, weight in keywords}

sentences = text.split('。')

sentence_importance = []

for sentence in sentences:

    words = jieba.lcut(sentence)

    importance = sum(keyword_dict.get(word, 0) for word in words)

    sentence_importance.append((sentence, importance))

sentences = sorted(sentence_importance, key=lambda x: x[1], reverse=True)

print("summary:")

print('\n'.join([i[0 ]for i in sorted_sentences[:3]]))