Detect and classify abandoned object on public space.

Abandoned Object Detection (AOD) systems are designed to automatically identify unattended or suspicious objects in public areas using computer vision and machine learning techniques. These systems help enhance public safety by providing real-time alerts and supporting security personnel in monitoring large spaces efficiently.

• Demo
• Stack
• Limitation

• Video and Frame Processing: OpenCV
• Classify Object: YOLOv11-classification

1. Clone The Repository

git clone http://github.com/Abandoned-Object-Detection
cd Abandoned-Object-Detection

2. Install Packages

pip install -r requirements.txt

Note:
The current project has been tested on Windows 11 + CUDA 11.8 installed with python version 3.10.11. If you have different version of CUDA, please refer to pytorch installation or previous versions

3. Video Configuration

video_choosen = "video3"
poly_used = POLY_ZONE_VIDEO3
file_path = "videos//video3.avi"
still_bg_path = "videos//video3.png"

Variables:

• video_choosen: Folder name where the output path.
• poly_used: Preset polygon coordinates defining the area of interest in the video. (You can create your own with PolygonZone from Roboflow)
• file_path: Path to the input video file.
• still_bg_path: Path to the reference background image used for comparison.

4. Run Program

python main.py

1. The algorithm does not adjust for changes in room lighting. This means the background image (ground truth) and the video stream must have the same lighting condition.
2. Good result: Video1, Video2, Video3, Video6-cut, Video7-cut, Video8-cut, Video9, and Video10
3. Bad result: Video4, Video5
4. Not included (have dynamic lighting condition): Video6, Video7, Video8

*-cut: mean that the video has been cut and trim to the scene that has good lighting condition