Annotation-Loop-Showcase/README.md

Example repository for a training loop based on Data Engine and YOLO

Using a COCO128 dataset provided by Ultralyltics

Repository structure

This repository showcases training a model on DagsHub using Data Engine as a store for annotations.

The repository has the same COCO128 dataset in two variants:

• All data in single folder
• Data split into a train and validation folder

The data and data_split folder contains the images and annotations.

The scripts and scripts_split folders contain python scripts for creating the Data Engine datasource, ingesting existing annotations into the datasource, then importing those annotations back and training model with the data.