YOLOv1-PyTorch

A deep dive into every component of the YOLO V1 object detector - from downloading and preparing the Pascal VOC data, through network architecture and loss formulation, all the way to training, evaluation, and real-time inference. This repository is designed to give you a clear, end-to-end understanding of how YOLO V1 works under the hood in PyTorch.

---

Repository Structure

YOLOv1-PyTorch/  
├── YOLO V1/                           # directory containing both notebooks  
│   ├── YOLO-V1-Explanation.ipynb      # highly comprehensive deep-dive notebook  
│   │   ├── Pascal VOC download & preprocessing  
│   │   ├── Detailed model architecture breakdown  
│   │   ├── Loss function derivation & implementation  
│   │   ├── Training loop with logging & checkpoints  
│   │   └── Inference examples with visualization  
│   └── YOLO-V1-Pure-Code.ipynb        # end-to-end pipeline: pure code only  
│       ├── Data loading  
│       ├── Model definition  
│       ├── Loss & optimization  
│       ├── Training & evaluation  
│       └── Inference with bounding-box visualization  
├── sheep.png                          # sample inference result on a sheep image  
├── biker.png                          # sample inference result on a bicycle image
├── LICENSE                            # MIT License text  
└── README.md                          # this document    

---

Notebooks Overview

• YOLO-V1-Explanation.ipynb
  A comprehensive, annotated walkthrough. Explains why each component exists, how it’s implemented in PyTorch, and how data flows through the network.

• YOLO-V1-Pure-Code.ipynb
  The identical implementation stripped of all commentary. Ideal for quick experiments, benchmarking, or integration into other projects.

---

Sample Inference Results

Below are two example outputs produced by the YOLO V1 pipeline:

---

Prerequisites

• Python 3.7+
• pip
• (Optional) CUDA-enabled GPU for accelerated training and inference

Install the required packages:

pip install torch torchvision opendatasets numpy matplotlib pillow tqdm

---

Getting Started

1. Clone the repository

git clone https://github.com/franciszekparma/YOLOv1-PyTorch.git
cd YOLOv1-PyTorch

2. Launch Jupyter Lab or Notebook

jupyter lab   # or: jupyter notebook

3. Open and run the notebooks • YOLO-V1-Explanation.ipynb for the full, annotated deep dive. • YOLO-V1-Pure-Code.ipynb for a no-frills, code-only run.

---

Contributing

Contributions are welcome. Please open an issue or submit a pull request for:
• Bug fixes
• Performance optimizations
• Support for additional datasets
• Documentation improvements

---

Acknowledgements

• “You Only Look Once” (YOLO V1) by Joseph Redmon et al.
• Aladdin Persson for providing Pascal VOC dataset via the opendatasets library.
• Inspiration and reference code from Aladdin Persson’s PyTorch tutorials.
• Tushar Kumar (ExplainingAI) for his theory-focused YOLO V1 tutorial.

---

License

This project is licensed under the MIT License.
© franciszekparma