YOLOv11 VisDrone Fine-Tuning and Evaluation

🛠️ Description

This repository contains a set of Python scripts that perform the following tasks:

1. Download the VisDrone dataset using the provided YAML configuration file from the Ultralytics YOLOv11 repository.
2. Fine-tune the YOLOv11s model on the VisDrone dataset.
3. Compare and evaluate the results before and after fine-tuning.

This setup is useful for anyone interested in experimenting with drone-based object detection using the latest YOLOv11 models.

⚙️ Languages or Frameworks Used

Main required modules:

• ultralytics
  • which includes torch, opencv, numpy and more

All dependencies are listed in pyproject.toml

Installation

Option 1: Using Poetry
This project uses Poetry for dependency management.
To install dependencies, first make sure Poetry is installed, then run:

poetry install

Option 2: Using pip

pip install $(poetry export --without-hashes --format=requirements.txt)

🌟 How to run

1️⃣ Download Dataset

• Run the datasets/download_visDrone_dataset.py

2️⃣ Fine-Tune YOLOv11s (Optional)

• Run the train_11s.py

3️⃣ Run Object Tracking

python3 main.py
or
poetry run python3 main.py

🪽 Results

• Train graphs over 100 epochs

• Confusion Matrix of the Newly Trained 11s model

🤖 Contributors

• kmsrogerkim
• Seongeun
• Doyeon
• Jeonghyun-Kang