MediaPipe生成带姿态关键点的视频无法打开问题求助

尝试保存MediaPipe生成的带姿态关键点的视频，终端无报错且文件已保存，但无法打开，报错信息：

播放失败：文件无法播放，可能格式不支持、扩展名错误或文件损坏，错误码0xc10100be
（德语原文：Wiedergabe nicht möglich，Die Datei kann nicht wiedergegeben werden. Das Dateiformat wird möglicherweise nicht unterstützt, die Dateierweiterung ist möglicherweise falsch oder die Datei ist beschädigt. 0xc10100be）

使用的代码如下：

import cv2
import mediapipe as mp
import numpy as np #for saving


mp_drawing = mp.solutions.drawing_utils
mp_drawing_styles = mp.solutions.drawing_styles
mp_pose = mp.solutions.pose

# Video Input (Replace 'your_video_file.mp4' with your video file path)
#cap = cv2.VideoCapture('Video_Reemt_Arm_1.mp4')
cap = cv2.VideoCapture('Reemt_Arms.mp4')
#cap = cv2.VideoCapture('Reemt_Feet.mp4')
#cap = cv2.VideoCapture('Reemt_Atmen_PTK.mp4') #dificil de detectar, intenta reconocimiento facial!
             
#cap = cv2.VideoCapture('Sam_Feet.mp4')
#cap = cv2.VideoCapture('Sam_Body.mp4')
#cap = cv2.VideoCapture('Sam_Atmen.mp4')

width = 600
height = 550

# Set the output video file name and parameters
output_file = 'output_video_with_pose.mp4'
fourcc = cv2.VideoWriter_fourcc(*'mp4v')  # Specify the codec (other options include 'XVID', 'MJPG', etc.)
out = cv2.VideoWriter('output_video.mp4', fourcc, 30.0, (width, height))

# Resizable window
cv2.namedWindow('Media Pipe Pose Estimation', cv2.WINDOW_NORMAL)

with mp_pose.Pose(
        min_detection_confidence=0.5,
        min_tracking_confidence=0.5) as pose:
    while cap.isOpened():
        success, image = cap.read()
        if not success:
            print("End of video")
            break

        # Drawing
        image.flags.writeable = False
        image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
        results = pose.process(image)

        image.flags.writeable = True
        image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)

        mp_drawing.draw_landmarks(
            image,
            results.pose_landmarks,
            mp_pose.POSE_CONNECTIONS,
            landmark_drawing_spec=mp_drawing_styles.get_default_pose_landmarks_style()
        )

        # Resize the window to match the video frame size
        cv2.resizeWindow('Media Pipe Pose Estimation', width, height)
        cv2.imshow('Media Pipe Pose Estimation', image)
        out.write(image)
        
        if cv2.waitKey(1) & 0xFF == 27:
            break

cap.release()
out.release() 
cv2.destroyAllWindows()

已尝试的方法：

• 更换VideoWriter编码（如XVID、MJPG、H264），示例代码：

fourcc = cv2.VideoWriter_fourcc(*'XVID')
output_file = 'output_video_with_pose.avi'  # Use .avi extension
out = cv2.VideoWriter(output_file, fourcc, 30.0, (width, height))

• 更新OpenCV
• 使用不同的媒体播放器
• 检查原视频
• 调试

核心问题出在输出视频参数与实际写入帧不匹配，具体如下：

1. 帧尺寸不匹配 #

你手动指定了width=600和height=550，但原视频的实际帧尺寸大概率不是这个数值，直接写入原尺寸帧到指定尺寸的VideoWriter中，会导致视频文件结构损坏。

正确做法是从原视频获取真实尺寸：

# 从原视频提取实际宽高
width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))

如果需要自定义输出尺寸，必须在写入前对每帧图像做resize处理：

image = cv2.resize(image, (target_width, target_height))
out.write(image)

2. 帧率不匹配 #

手动设置的30.0帧率可能与原视频不一致，同样会导致播放异常，应从原视频获取真实帧率：

fps = cap.get(cv2.CAP_PROP_FPS)

修正后的完整代码 #

import cv2
import mediapipe as mp

mp_drawing = mp.solutions.drawing_utils
mp_drawing_styles = mp.solutions.drawing_styles
mp_pose = mp.solutions.pose

# 读取原视频
cap = cv2.VideoCapture('Reemt_Arms.mp4')

# 获取原视频的真实宽、高、帧率
width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
fps = cap.get(cv2.CAP_PROP_FPS)

# 设置输出参数，确保编码与扩展名匹配
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
out = cv2.VideoWriter('output_video_with_pose.mp4', fourcc, fps, (width, height))

cv2.namedWindow('Media Pipe Pose Estimation', cv2.WINDOW_NORMAL)

with mp_pose.Pose(
        min_detection_confidence=0.5,
        min_tracking_confidence=0.5) as pose:
    while cap.isOpened():
        success, image = cap.read()
        if not success:
            print("End of video")
            break

        image.flags.writeable = False
        image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
        results = pose.process(image)

        image.flags.writeable = True
        image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)

        mp_drawing.draw_landmarks(
            image,
            results.pose_landmarks,
            mp_pose.POSE_CONNECTIONS,
            landmark_drawing_spec=mp_drawing_styles.get_default_pose_landmarks_style()
        )

        # 写入帧（确保尺寸与VideoWriter一致）
        out.write(image)

        cv2.resizeWindow('Media Pipe Pose Estimation', width, height)
        cv2.imshow('Media Pipe Pose Estimation', image)
        
        if cv2.waitKey(1) & 0xFF == 27:
            break

cap.release()
out.release() 
cv2.destroyAllWindows()

额外注意事项 #

• 编码与扩展名必须匹配：mp4v对应.mp4，XVID对应.avi，MJPG对应.avi或.mp4
• 若仍无法播放，可尝试cv2.VideoWriter_fourcc(*'avc1')（H.264编码），需确保OpenCV支持该编码
• 必须调用out.release()完成文件写入，否则文件会因未收尾而损坏

内容的提问来源于stack exchange，提问作者user22801597