COCO 1K

A Dataset of ~1,199 images (1,100 training and 99 validation) from the COCO-2017 dataset. Created for the purposes of fast dataset iteration and prototyping. Demos the usage of DagsHub Data Engine which enables you to iterate on your datasets quickly.

To use with the DagsHub Data Engine:

pip install dagshub

And then in your Python:

from dagshub.data_engine import datasources

ds = datasources.get('Dean/COCO_1K', 'COCO_1K')

ds.head().dataframe

To learn more about Data Engine, check out the Docs

You can use the included Notebook to create an end to end flow of the Data Engine, which will enable you to create the datasource, query it, visualize it, train a model, add new data and annotate it using a machine learning model backend, and improve the model performance by improving the data quality.