Chiller-V8-Pipeline

This project implements an MLOps pipeline for YOLOv8-based chiller detection.
It uses DVC for dataset and model versioning, AWS S3 for remote storage, and GitHub Actions for CI/CD automation.

Features

• YOLOv8 object detection training
• Data and model versioning with DVC
• Remote storage on AWS S3 (chiller-dvc-pipeline)
• Automated workflows (train, validate, upload artifacts)
• TensorBoard support for logs and metrics

Project Structure

chiller-v8-pipeline/ │── data/ # Dataset folder │── models/ # Model weights (tracked by DVC) │── notebooks/ # Jupyter notebooks │── .github/workflows/ # CI/CD workflows │── requirements.txt # Dependencies

How to Run

1. Clone repo & install dependencies

   git clone https://github.com/<your-username>/chiller-v8-pipeline.git
   cd chiller-v8-pipeline
   pip install -r requirements.txt
   
    2.	Pull dataset and model
   

dvc pull

3.	Run training

yolo task=detect mode=train data=chillertrain.yaml model=yolov8n.pt epochs=50 imgsz=640

CI/CD • Validation workflow automatically pulls the latest model from DVC. • Training workflow can be extended for retraining and deployment.

⸻

🚀 End-to-end reproducible ML pipeline with YOLOv8 + DVC + GitHub Actions + AWS S3.