预测图像中的点序列

import numpy as np
import cv2
import os

def load_data(image_dir, points_file):
    images = []
    points = []
    
    with open(points_file, 'r') as f:
        for line in f:
            parts = line.strip().split(',')
            image_path = os.path.join(image_dir, parts[0])
            image = cv2.imread(image_path)
            image = cv2.resize(image, (128, 128))  # 调整图像大小
            images.append(image)
            
            point_sequence = np.array([float(p) for p in parts[1:]])
            points.append(point_sequence)
    
    return np.array(images), np.array(points)

image_dir = 'path/to/images'
points_file = 'path/to/points.txt'
images, points = load_data(image_dir, points_file)

import tensorflow as tf
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input, Conv2D, MaxPooling2D, Flatten, LSTM, Dense

def build_model(input_shape, output_size):
    inputs = Input(shape=input_shape)
    
    # CNN 部分
    x = Conv2D(32, (3, 3), activation='relu')(inputs)
    x = MaxPooling2D((2, 2))(x)
    x = Conv2D(64, (3, 3), activation='relu')(x)
    x = MaxPooling2D((2, 2))(x)
    x = Flatten()(x)
    
    # RNN 部分
    x = tf.expand_dims(x, axis=1)  # 添加时间步维度
    x = LSTM(128, return_sequences=True)(x)
    x = LSTM(128)(x)
    
    # 输出层
    outputs = Dense(output_size)(x)
    
    model = Model(inputs, outputs)
    return model

input_shape = (128, 128, 3)
output_size = points.shape[1]
model = build_model(input_shape, output_size)
model.compile(optimizer='adam', loss='mse')
model.summary()

from sklearn.model_selection import train_test_split

X_train, X_val, y_train, y_val = train_test_split(images, points, test_size=0.2, random_state=42)

history = model.fit(X_train, y_train, epochs=50, batch_size=32, validation_data=(X_val, y_val))

# 评估模型
loss = model.evaluate(X_val, y_val)
print(f'Validation Loss: {loss}')

# 使用模型进行预测
sample_image = X_val[0]
predicted_points = model.predict(np.expand_dims(sample_image, axis=0))
print(f'Predicted Points: {predicted_points}')