Python结合OBS WebSocket人脸检测场景切换异常排查求助

我正在尝试基于摄像头画面的人脸检测结果，通过Python调用OBS WebSocket API自动切换OBS场景。目前遇到的问题是：即使检测到人脸，场景也无法切换；程序无报错，但调试输出显示GetCurrentScene()始终返回空字典{}，且无法正常获取OBS WebSocket版本信息。

import cv2
import time
import numpy as np
from obswebsocket import obsws, requests, events

# OBS WebSocket connection settings
host = "localhost"
port = 4455
password = "PLACEHOLDER"

# Scene names
live_scene = "LIVE"
afk_scene = "AFK"

# Global variable to store the current scene name
current_scene_name = ""

# Connect to OBS
ws = obsws(host, port, password)
print("Connecting to OBS WebSocket at {}:{}".format(host, port))
ws.connect()

try:
    version_info = ws.call(requests.GetVersion())
    print("Connected to OBS WebSocket version: {}.{}.{}".format(
        version_info.data().get("majorVersion", "Unknown"),
        version_info.data().get("minorVersion", "Unknown"),
        version_info.data().get("patchVersion", "Unknown")
    ))
    scenes = ws.call(requests.GetSceneList())
    print("Available scenes:", [scene["name"] for scene in scenes.data()["scenes"]])  # Added debug output
except Exception as e:
    print("Error connecting to OBS WebSocket:")
    print(type(e))
    print(e.args)
    print(e)

# Initialize webcam
cap = cv2.VideoCapture(0)

# Initialize face detection
face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')

while True:
    ret, frame = cap.read()

    if ret:
        gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)

        faces = face_cascade.detectMultiScale(gray, 1.1, 4)

        if len(faces) > 0:
            print("Face detected")
            current_scene = ws.call(requests.GetCurrentScene())
            print("Current scene:", current_scene.data())  # Added debug output
            current_scene_name = current_scene.data().get("name", "Unknown")
            ws.call(requests.SetCurrentScene(scene_name=live_scene))
            print("Switching to scene:", live_scene)
        else:
            print("No face detected")
            current_scene = ws.call(requests.GetCurrentScene())
            print("Current scene:", current_scene.data())  # Added debug output
            current_scene_name = current_scene.data().get("name", "Unknown")
            ws.call(requests.SetCurrentScene(scene_name=afk_scene))
            print("Switching to scene:", afk_scene)

        time.sleep(1)

    else:
        print("Error reading frame from webcam")
        break

cap.release()
cv2.destroyAllWindows()

Connected to OBS WebSocket version: Unknown.Unknown.Unknown
Current scene: {}
Switching to scene: LIVE

核心原因 #

你使用的obswebsocket库仅兼容OBS WebSocket 4.x版本，但当前OBS默认安装的是5.x版本，两者API协议完全不兼容。旧库无法解析新版本的响应格式，因此出现返回空字典、无法获取版本信息的情况。

具体解决步骤 #

1. 替换兼容5.x版本的库
   卸载旧库，安装官方推荐的适配5.x版本的obs-websocket-py：

   pip uninstall obswebsocket
   pip install obs-websocket-py
   

2. 修改代码适配新API
   新库的请求方法、数据结构均有变化，以下是调整后的完整代码：

   import cv2
   import time
   from obswebsocket import obsws, requests
   
   # OBS WebSocket 5.x连接配置
   host = "localhost"
   port = 4455
   password = "PLACEHOLDER"
   
   live_scene = "LIVE"
   afk_scene = "AFK"
   
   # 建立连接
   ws = obsws(host, port, password)
   ws.connect()
   
   try:
       # 获取版本信息（新API方法）
       version_resp = ws.call(requests.GetVersion())
       version_data = version_resp.get_data()
       print(f"Connected to OBS WebSocket version: {version_data['obsWebSocketVersion']}")
       
       # 获取场景列表（新API方法）
       scenes_resp = ws.call(requests.GetSceneList())
       scenes = scenes_resp.get_data()["scenes"]
       print("Available scenes:", [scene["sceneName"] for scene in scenes])
   except Exception as e:
       print(f"Error connecting to OBS WebSocket: {e}")
   
   # 初始化摄像头与人脸检测
   cap = cv2.VideoCapture(0)
   face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')
   
   while True:
       ret, frame = cap.read()
       if not ret:
           print("Error reading frame from webcam")
           break
   
       gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
       faces = face_cascade.detectMultiScale(gray, 1.1, 4)
   
       target_scene = live_scene if len(faces) > 0 else afk_scene
       
       # 获取当前程序场景（新API方法）
       current_scene_resp = ws.call(requests.GetCurrentProgramScene())
       current_scene = current_scene_resp.get_data()["currentProgramSceneName"]
       
       # 仅当场景不一致时切换，减少无效请求
       if current_scene != target_scene:
           ws.call(requests.SetCurrentProgramScene(sceneName=target_scene))
           print(f"Switched to scene: {target_scene}")
       else:
           print(f"Already in scene: {target_scene}")
   
       time.sleep(1)
   
   # 释放资源
   cap.release()
   cv2.destroyAllWindows()
   ws.disconnect()
   

3. 检查OBS WebSocket设置

   • 打开OBS，进入工具 > obs-websocket设置
   • 确认端口为4455，密码与代码中PLACEHOLDER一致
   • 勾选启用Web服务器

4. 额外优化建议

   • 增加连续帧人脸检测判断（比如连续2帧检测到人脸再切换），避免误触发
   • 程序结束时调用ws.disconnect()关闭连接，避免资源泄漏

内容的提问来源于stack exchange，提问作者Nes Elm