CNN-LSTM模型构建和训练

Question

我正在尝试从视频数据构建 cnn-lstm 模型。这是特征数据frames_data=(150120, 128, 128, 3)和label_data=(150120, 19)的形状，我如何重塑它们以用作cnn-lstm模型的输入。谢谢 这是提取帧的代码

class_names = os.listdir(root_folder_path)
    
    # Empty arrays to store extracted frames and corresponding labels
    frames_data = []
    labels_data = []
    
    # Function to extract frames from a video file, normalize them, and append to the arrays
    def extract_frames(video_path, label, num_frames=10, frame_size=(128, 128)):
        cap = cv2.VideoCapture(video_path)
        frame_count = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
        frame_indices = np.linspace(0, frame_count - 1, num_frames, dtype=np.int)
        frame_num = 0
        while True:
            ret, frame = cap.read()
            if not ret:
                break
            if frame_num in frame_indices:
                # Resize the frame to the desired size
                resized_frame = cv2.resize(frame, frame_size)
                # Normalize the frame to values between 0 and 1
                normalized_frame = resized_frame / 255.0
                frames_data.append(normalized_frame)
                labels_data.append(label)
            frame_num += 1
        cap.release()
    
    # Parameters for frame extraction
    num_frames_per_video = 10
    frame_size = (128, 128)
    
    # Loop through each class folder and extract frames
    for i, class_name in enumerate(class_names):
        class_folder_path = os.path.join(root_folder_path, class_name)
        video_files = [f for f in os.listdir(class_folder_path) if f.endswith('.mp4')]
        for video_file in video_files:
            video_path = os.path.join(class_folder_path, video_file)
            extract_frames(video_path, i, num_frames=num_frames_per_video, frame_size=frame_size)
    
    # Convert lists to numpy arrays for easier manipulation (optional)
    frames_data = np.array(frames_data)
    labels_data = np.array(labels_data) 

我尝试重塑它，但它给了我错误

ValueError: Input 0 of layer "max_pooling2d_16" is incompatible with the layer: expected ndim=4, found ndim=5. Full shape received: (None, 10, 128, 128, 16)

---