Python-VOT: Visual Object Tracking Challenge 👁️📦

Python-VOT is a Computer Vision project focused on the development and evaluation of an object tracking algorithm. The project is built upon the VOT (Visual Object Tracking) Challenge framework, specifically using datasets and metrics from the VOT 2017 edition.

The goal is to accurately estimate the bounding box of a target object across various image sequences, handling challenges such as motion blur, illumination changes, and occlusions.

🚀 Project Overview

This implementation was developed as part of the Robotics and Artificial Vision course at Universidad Pablo de Olavide. It follows the official VOT methodology to compare tracking performance against ground truth data.

Key Metrics Evaluated:

• Accuracy: The average overlap between the predicted and ground truth bounding boxes.
• Robustness: The number of times the tracker fails (loses the object) during a sequence.
• EAO (Expected Average Overlap): A primary ranking metric that combines both accuracy and robustness.

📁 Repository Structure

The project is organized around the official VOT Toolkit workspace:

• /results: Contains the final source code (vision_challenge_tracker.py) and the consolidated performance report.
• /vot-workspace: The core environment for testing and analysis.
  • tracker/: Contains the implementation of the tracker and the vot.py integration API.
  • sequences/: Includes the benchmark video sequences (e.g., ball1, basketball, bolt1, car1, tiger).
  • analysis/: Timestamped folders containing detailed HTML reports for each evaluation run.
  • logs/: Execution logs for debugging tracker behavior.
• Vision_Challenge-english-final.pdf: The official challenge documentation and requirements.

🛠️ Tech Stack

• Language: Python.
• Framework: VOT Toolkit.
• Libraries: Likely utilizing OpenCV, NumPy, and Matplotlib for image processing and visualization.

📊 Evaluation Workflow

To evaluate the tracker, the following toolkit commands are used:

1. Test: Verify the tracker integration.

   vot test VCTracker

2. Evaluate: Run the tracker on all provided sequences.

   vot evaluate VCTracker

3. Analysis: Generate the performance reports in HTML format.

   vot analysis VCTracker

👤 Author

• Luis Carmona - eLeCe2611

---

Developed for academic purposes to explore state-of-the-art Visual Object Tracking techniques.