### 修正 `ValueError: Cannot reshape array of size 4 into shape (112, 92)` 并实现基于 ORL 数据集的人脸识别代码 以下是一个完整的解决方案，包括修正错误和实现基于 ORL 数据集的人脸识别的步骤。 --- #### 1. 修正 `ValueError` 错误 该错误的原因是尝试将一个大小为 4 的数组重塑为形状 `(112, 92)`，但目标形状的元素总数（112 * 92 = 10304）与原始数组的元素总数（4）不匹配[^2]。以下是修正方法： ```python import numpy as np # 示例数据 array = np.array([1, 2, 3, 4]) # 检查数组大小是否匹配目标形状 if array.size != 112 * 92: # 如果不匹配，尝试填充或裁剪数组以适配目标形状 target_shape = (112, 92) adjusted_array = np.zeros(target_shape, dtype=array.dtype) rows_overlap = min(array.size // 92, target_shape[0]) cols_overlap = min(92, target_shape[1]) adjusted_array[:rows_overlap, :cols_overlap] = array[:rows_overlap * cols_overlap].reshape(rows_overlap, cols_overlap) else: adjusted_array = array.reshape(112, 92) print(f"Adjusted shape: {adjusted_array.shape}") ``` --- #### 2. 加载 ORL 数据集 ORL 数据集包含 400 张灰度图像，每张图像的尺寸为 92×112 像素。可以使用 `fetch_olivetti_faces` 函数从 `sklearn.datasets` 中加载数据[^1]。 ```python from sklearn.datasets import fetch_olivetti_faces # 加载 ORL 数据集 data = fetch_olivetti_faces() X = data.images # 图像数据，形状为 (n_samples, 64, 64) y = data.target # 标签数据，形状为 (n_samples,) ``` --- #### 3. 数据预处理 在加载数据后，需要对数据进行预处理，确保其形状适合后续操作。 ```python # 验证数据形状 print(f"Images shape: {X.shape}") # 输出应为 (400, 64, 64) print(f"Labels shape: {y.shape}") # 输出应为 (400,) # 调整图像大小为 (112, 92) import cv2 def resize_images(images, target_shape): resized_images = [] for image in images: resized_image = cv2.resize(image, (target_shape[1], target_shape[0])) resized_images.append(resized_image) return np.array(resized_images) X_resized = resize_images(X, (112, 92)) print(f"Resized images shape: {X_resized.shape}") # 输出应为 (400, 112, 92) ``` --- #### 4. 训练人脸识别模型 可以使用主成分分析（PCA）降维和最近邻分类器（KNN）来实现人脸识别。 ```python from sklearn.model_selection import train_test_split from sklearn.decomposition import PCA from sklearn.neighbors import KNeighborsClassifier from sklearn.metrics import accuracy_score # 划分训练集和测试集 X_train, X_test, y_train, y_test = train_test_split( X_resized.reshape(len(X_resized), -1), y, test_size=0.2, random_state=42 ) # 应用 PCA 降维 pca = PCA(n_components=150) # 选择保留 150 个主成分 X_train_pca = pca.fit_transform(X_train) X_test_pca = pca.transform(X_test) # 训练 KNN 分类器 knn = KNeighborsClassifier(n_neighbors=5) knn.fit(X_train_pca, y_train) # 测试模型性能 y_pred = knn.predict(X_test_pca) accuracy = accuracy_score(y_test, y_pred) print(f"Accuracy: {accuracy:.2f}") ``` --- #### 5. 可视化结果 可以通过可视化来展示识别结果。 ```python import matplotlib.pyplot as plt # 显示测试集中第一张图像及其预测结果 idx = 0 plt.imshow(X_test[idx].reshape(112, 92), cmap='gray') plt.title(f"Predicted: {y_pred[idx]}, True: {y_test[idx]}") plt.axis('off') plt.show() ``` --- ###