Artificial Vision Group 1 - Person Tracking and Attribute Recognition

This project implements a system for tracking people in video streams, detecting line crossings, and recognizing attributes such as gender, bag presence, and hat presence. It utilizes YOLO for object detection and a custom Multi-Task CNN for attribute recognition (PAR).

Description

The tracker.py script processes video files to:

1. Track People: Uses YOLOv8 (or YOLOv8-seg) and BoT-SORT for robust person tracking.
2. Detect Line Crossings: Counts people crossing defined lines in specific directions.
3. Recognize Attributes: Classifies detected people for:
   • Gender: Male / Female
   • Bag: Yes / No
   • Hat: Yes / No

Libraries Used

The project relies on the following key libraries:

• Ultralytics YOLO: For object detection and tracking.
• PyTorch: Deep learning framework for the PAR model.
• OpenCV (cv2): Video processing and drawing visualizations.
• NumPy: Numerical operations.
• Shapely: Geometric operations for polygon containment and line intersections.
• Pillow (PIL): Image manipulation.

Usage

To run the tracker on a video file:

python tracker.py --input <path_to_video> --output <path_to_output> --config <path_to_config>

Arguments

Argument	Default	Description
--input	../video.mp4	Path to the input video file.
--output	./output/output.mp4	Path where the processed video will be saved.
--config	./config/camera.json	Path to the camera configuration file (defines lines and zones).

Example

python tracker.py --input data/my_video.mp4 --output results/tracked_video.mp4

Alternatively, you can run the script with the default arguments:

python tracker.py

Authors

• Pietro Martano
• Angelo Molinario
• Massimiliano Ranauro
• Antonio Sessa