lzy/sci-gui-agent-benchmark
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

VLC updates, and some infra bugs fix

Browse Source
This commit is contained in:
Timothyxxx
2024-01-09 23:14:06 +08:00
parent 457a0498f2
commit abcafce750
10 changed files with 291 additions and 121 deletions
Download Patch File Download Diff File Expand all files Collapse all files

72
desktop_env/evaluators/metrics/vlc.py
View File

@@ -7,8 +7,6 @@ from xml.etree import ElementTree
import acoustid
import cv2
import imagehash
import pyautogui
import pygetwindow as gw # todo: change to the library that supports Linux
from PIL import Image
logger = logging.getLogger("desktopenv.metrics.vlc")
@@ -72,30 +70,11 @@ def is_vlc_recordings_folder(actual_config_path: str, rule: Dict[str, str]) -> f
return False
def are_audio_files_similar(mp3_file_path, mp4_file_path):
# Extract audio fingerprint from MP3 file
mp3_fingerprint, mp3_duration = acoustid.fingerprint_file(mp3_file_path)
# Extract the audio stream from the MP4 file
mp4_audio_path = os.path.splitext(mp4_file_path)[0] + '_extracted.mp3'
try:
subprocess.run(["ffmpeg", "-i", mp4_file_path, "-vn", "-ar", "44100", "-ac", "2", "-ab", "192k", "-f", "mp3",
mp4_audio_path], check=True)
except subprocess.CalledProcessError as e:
print(f"An error occurred during audio extraction from MP4: {e}")
return False
# Extract audio fingerprint from the extracted audio
mp4_fingerprint, mp4_duration = acoustid.fingerprint_file(mp4_audio_path)
# Clean up temporary extracted audio file
os.remove(mp4_audio_path)
# Compare fingerprints (rudimentary comparison)
if mp3_duration >= mp4_duration and mp3_fingerprint == mp4_fingerprint:
def is_vlc_fullscreen(actual_window_size, screen_size):
if actual_window_size['width'] == screen_size['width'] and actual_window_size['height'] == screen_size['height']:
return True
return False
else:
return False
def compare_videos(video_path1, video_path2, max_frames_to_check=100, threshold=5):
@@ -137,28 +116,27 @@ def compare_videos(video_path1, video_path2, max_frames_to_check=100, threshold=
return True
def is_vlc_fullscreen():
"""
Checks if the VLC window is in full-screen mode.
def are_audio_files_similar(mp3_file_path, mp4_file_path):
# Extract audio fingerprint from MP3 file
mp3_fingerprint, mp3_duration = acoustid.fingerprint_file(mp3_file_path)
When VLC is in full-screen mode, its window size matches the screen size with no borders.
"""
# Extract the audio stream from the MP4 file
mp4_audio_path = os.path.splitext(mp4_file_path)[0] + '_extracted.mp3'
try:
# Get the VLC window; adjust the title as per your VLC window's title
vlc_window = gw.getWindowsWithTitle('VLC media player')[0] # Adjust title if needed
if not vlc_window:
return False
# Get screen size
screen_width, screen_height = pyautogui.size()
# Check if VLC window size matches the screen size
return (vlc_window.width == screen_width and vlc_window.height == screen_height)
except IndexError:
# VLC window not found
logger.error("VLC window not found.")
return False
except Exception as e:
logger.error(f"An error occurred: {e}")
subprocess.run(["ffmpeg", "-i", mp4_file_path, "-vn", "-ar", "44100", "-ac", "2", "-ab", "192k", "-f", "mp3",
mp4_audio_path], check=True)
except subprocess.CalledProcessError as e:
print(f"An error occurred during audio extraction from MP4: {e}")
return False
# Extract audio fingerprint from the extracted audio
mp4_fingerprint, mp4_duration = acoustid.fingerprint_file(mp4_audio_path)
# Clean up temporary extracted audio file
os.remove(mp4_audio_path)
# Compare fingerprints (rudimentary comparison)
if mp3_duration >= mp4_duration and mp3_fingerprint == mp4_fingerprint:
return True
return False