feat: use mss instead of imagegrab

pyproject.toml

@@ -42,6 +42,7 @@ dependencies = [
     "timm",
     "einops",
     "modelscope",
+    "mss",
 ]
 
 [project.urls]

screen_recorder/common.py

@@ -4,11 +4,11 @@ import time
 import logging
 import platform
 import subprocess
-from PIL import Image, ImageGrab
+from PIL import Image
 import imagehash
 from memos.utils import write_image_metadata
-from screeninfo import get_monitors
 import ctypes
+from mss import mss
 
 if platform.system() == "Windows":
     import win32gui
@@ -173,25 +173,17 @@ def take_screenshot_windows(
     app_name,
     window_title,
 ):
-    for monitor in get_monitors():
-        safe_monitor_name = "".join(
-            c for c in monitor.name if c.isalnum() or c in ("_", "-")
-        )
+    with mss() as sct:
+        for i, monitor in enumerate(sct.monitors[1:], 1):  # Skip the first monitor (entire screen)
+            safe_monitor_name = f"monitor_{i}"
             logging.info(f"Processing monitor: {safe_monitor_name}")
 
             webp_filename = os.path.join(
                 base_dir, date, f"screenshot-{timestamp}-of-{safe_monitor_name}.webp"
             )
 
-        img = ImageGrab.grab(
-            bbox=(
-                monitor.x,
-                monitor.y,
-                monitor.x + monitor.width,
-                monitor.y + monitor.height,
-            )
-        )
-        img = img.convert("RGB")
+            img = sct.grab(monitor)
+            img = Image.frombytes("RGB", img.size, img.bgra, "raw", "BGRX")
             current_hash = str(imagehash.phash(img))
 
             if (