Pet Tracking Project

This project implements pet tracking using an OpenCV2-based algorithm combined with a YOLO model. The application is designed to analyze video frames and track pets dynamically.

---

Environment Setup

Requirements

• Conda version: 3
• Python version: 3.10
• OpenCV version: 4.5.5

Setup Instructions

1. Create a Conda environment:

   conda env create -f enviroment.yml -n net_tracking

2. Export your own environment:

   conda env export > environment.yml 

3. Clone the YOLO model and set up the classification ID file:
   • YOLO Model Classification IDs

---

How to Run

Executable Script

The main script for running the pet tracking project is tracker_test.exe. This is a compiled executable built with PyInstaller for ease of use.

Command Structure

.\dist\tracker_test.exe <video_path> <model_class_id> <model_confidence> -d <debug_mode> -s <strategy> --minInterval <min_time> --maxInterval <max_time> --strategyInterval <strategy_time> --resize <width> <height>

Example Command

.\dist\tracker_test.exe D:\video.mp4 15 0.1 -d True -s WHEN_LOST --minInterval 500 --maxInterval 0 --strategyInterval 0 --resize 1920 1080

Parameter Descriptions

Parameter	Description	Example
<video_path>	Path to the input video file.	D:\video.mp4
<model_class_id>	The class id of the model match	15
<model_confidence>	Minimum confidence of the model	0.1
-d <debug_mode>	Debug mode (True/False).	True
-s <strategy>	Tracking strategy (e.g., WHEN_LOST, WHEN_FREE, etc.).	WHEN_LOST
--minInterval	Minimum time interval (in milliseconds).	500
--maxInterval	Maximum time interval (in milliseconds).	0
--strategyInterval	Interval for applying strategy (in milliseconds).	0
--resize	Scale the video to the target size.	1920 1080
Use the .\dist\tracker_test.exe --help command to know more

---

YOLO Model Information

The YOLO model is used for pet detection and classification. The classification IDs are defined in the file:

• coco.names

You can find YOLO models at the following web sites:

• YOLO4

Ensure that this file is downloaded and correctly referenced in your project.

---

Compile and pack

This project uses pyInstaller for packaging, the following is the packaging command

For Windows

pyinstaller --onefile --add-binary "~\env\conda\envs\my_project_env\Library\bin\*.dll;." .\tests\csrt\tracker_test.py

For Linux

pyinstaller --onefile --add-binary "~/env/conda/envs/my_project_env/lib/*.so;." .\tests\csrt\tracker_test.py

---

Notes and Troubleshooting

1. Environment Compatibility: This project is designed for Python 3.10 and OpenCV 4.5.5. Ensure your environment matches these versions.
2. Debug Mode: If you encounter unexpected behavior, enable debug mode (-d True) to log additional details, This outputs the processed video stream.
3. Model Dependencies: Make sure all YOLO dependencies and weights are properly downloaded and configured.
4. Performance: Adjust parameters like max_distance and iou_threshold to optimize tracking performance for your specific video input.

---

License

This project is licensed under the MIT License - see the LICENSE file for details.

---

For additional information or issues, feel free to contact us or raise an issue in the project repository.