A real-time people detection system that counts individuals in webcam feeds or uploaded videos. Built for analyzing vlog footage and timelapse videos.
- Live webcam feed processing
- Video file upload and processing
- Real-time people counting (displayed at the top of the screen)
- Bounding box detection around entire bodies
- Frame-by-frame analysis
- Optimized for vlog footage and timelapse clips
This tool works particularly well for:
- Analyzing vlog footage to see crowd density
- Processing time-lapse videos to track people flow
- Counting participants in recorded events
- Live webcam monitoring
- Python 3.x
- OpenCV (
cv2) - Ultralytics YOLO
- PIL (Pillow)
- NumPy
pip install opencv-python ultralytics pillow numpyThe program will ask you to choose between:
- Live webcam mode - Uses your computer's webcam for real-time detection
- Video upload mode - Allows you to select and process a video file
Follow the on-screen instructions to select your preferred mode.
When using the video upload mode, please allow for the entire video to play through before closing the software. If you do not do this, the video will be saved at the time you close the program. For example If you were to upload a 50-second video and close the program after 10 seconds. Only the first 10 seconds will be saved to your local device
Actively working on improving detection accuracy, particularly for:
- Varying lighting conditions
- Multiple people in frame
- Different camera angles and distances
Check out the fac-rec branch for facial recognition capabilities (in development).