TensorFlow Keras模型预测时输入维度不匹配问题解析与解决方案

import numpy as np
import cv2
import tensorflow as tf
from tensorflow.keras import layers
from tensorflow.keras.models import Sequential

# 假设您的模型已经定义并加载
# 为了演示，我们定义一个简化的模型结构
img_height = 180
img_width = 180
channels = 3
num_classes = 10 # 示例值

model = Sequential([
    layers.Rescaling(1./255, input_shape=(img_height, img_width, channels)), # 也可以在这里定义input_shape
    layers.Conv2D(16, 3, padding='same', activation='relu'),
    layers.MaxPooling2D(),
    layers.Flatten(),
    layers.Dense(num_classes)
])
model.compile(optimizer='adam',
              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
              metrics=['accuracy'])

# 假设您已加载并预处理了图像
image_path = "C:\anImage\c000b634560ef3c9211cbf9e08ebce74.jpg"
image = cv2.imread(image_path)
image = cv2.resize(image, (img_width, img_height))
image = np.asarray(image).astype('float32')

print(f"原始图像维度: {image.shape}") # 输出 (180, 180, 3)

# 关键步骤：添加批次维度
# 方法一：使用 np.expand_dims
image_with_batch_dim = np.expand_dims(image, axis=0)
print(f"添加批次维度后图像维度 (np.expand_dims): {image_with_batch_dim.shape}") # 输出 (1, 180, 180, 3)

# 方法二：使用 np.newaxis
# image_with_batch_dim = image[np.newaxis, ...]
# print(f"添加批次维度后图像维度 (np.newaxis): {image_with_batch_dim.shape}")

# 进行预测
predictions = model.predict(image_with_batch_dim)
print("预测成功！")
print(f"预测结果形状: {predictions.shape}")

import tensorflow as tf
from tensorflow.keras import layers
from tensorflow.keras.models import Sequential

img_height = 180
img_width = 180
channels = 3
num_classes = 10 # 示例值

# 显式定义 InputLayer
model = Sequential([
    layers.InputLayer(input_shape=(img_height, img_width, channels)), # 明确指定输入形状
    layers.Rescaling(1./255),
    layers.Conv2D(16, 3, padding='same', activation='relu'),
    layers.MaxPooling2D(),
    layers.Conv2D(32, 3, padding='same', activation='relu'),
    layers.MaxPooling2D(),
    layers.Conv2D(64, 3, padding='same', activation='relu'),
    layers.MaxPooling2D(),
    layers.Flatten(),
    layers.Dense(128, activation='relu'),
    layers.Dense(num_classes)
])

model.compile(optimizer='adam',
              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
              metrics=['accuracy'])

model.summary() # 此时 summary 会显示完整的输入/输出形状

import matplotlib.pyplot as plt
import numpy as np
import os
import PIL
import tensorflow as tf
import cv2

from tensorflow import keras
from tensorflow.keras import layers
from tensorflow.keras.models import Sequential
import pathlib

# 定义图像尺寸和通道数
img_height = 180
img_width = 180
channels = 3

# 模拟数据加载和模型训练（仅为演示，实际训练过程更复杂）
# 假设您已经有了 train_ds 和 val_ds
# 这里为了代码可运行，简单模拟 num_classes
num_classes = 5 # 假设有5个类别

# 构建模型：显式定义 InputLayer 是一个好的实践
model = Sequential([
    layers.InputLayer(input_shape=(img_height, img_width, channels)), # 明确指定输入形状
    layers.Rescaling(1./255),
    layers.Conv2D(16, 3, padding='same', activation='relu'),
    layers.MaxPooling2D(),
    layers.Conv2D(32, 3, padding='same', activation='relu'),
    layers.MaxPooling2D(),
    layers.Conv2D(64, 3, padding='same', activation='relu'),
    layers.MaxPooling2D(),
    layers.Flatten(),
    layers.Dense(128, activation='relu'),
    layers.Dense(num_classes)
])

model.compile(optimizer='adam',
              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
              metrics=['accuracy'])

model.summary()

# 模拟模型训练（在实际应用中，您会用 train_ds 和 val_ds 进行训练）
# model.fit(train_ds, validation_data=val_ds, epochs=epochs)

# 准备单张图像进行预测
image_path = "C:\anImage\c000b634560ef3c9211cbf9e08ebce74.jpg" # 替换为您的图像路径
image = cv2.imread(image_path)

if image is None:
    print(f"错误：无法读取图像 {image_path}。请检查路径和文件是否存在。")
else:
    # 调整图像大小以匹配模型输入
    image = cv2.resize(image, (img_width, img_height))
    # 将图像数据转换为浮点型 numpy 数组
    image = np.asarray(image).astype('float32')

    print(f"原始图像维度: {image.shape}") # 应为 (180, 180, 3)

    # 关键步骤：为单张图像添加批次维度
    # 模型期望 (batch_size, H, W, C)，所以需要将 (H, W, C) 变为 (1, H, W, C)
    image_for_prediction = np.expand_dims(image, axis=0)
    print(f"用于预测的图像维度: {image_for_prediction.shape}") # 应为 (1, 180, 180, 3)

    # 进行预测
    try:
        predictions = model.predict(image_for_prediction)
        print("模型预测成功！")
        print(f"预测结果形状: {predictions.shape}")
        # 如果需要，可以进一步处理预测结果，例如：
        # predicted_class = np.argmax(predictions[0])
        # print(f"预测类别索引: {predicted_class}")
    except Exception as e:
        print(f"预测过程中发生错误: {e}")