By team ROUND https://round-ai.github.io/

Goals

We aim to help researchers and developers can maximize the usage of AI datasets and exploit most out of them. ROUND is an open-source data preparation service that helps create preprocess pipelines and visualize intermediate outputs very easily.

News

• 2024.08.04 Initial release!
• (ETA) 2024.08.10 GUI features will be released!

WORK IN PROGRESS

Installation

To install, use pip install -e .

How to Use

Say you have your own dataset under some dataset directory="./Dataset", you can first analyze the dataset using following code.

import round
round.analyze_dataset(dataset_pth="./Dataset")

Then, create a config file using the GUI we provide (Coming Soon!),

or manually like below.

You should create LOAD-00x.py, SPLIT-00x.py, PREPROCESS-00x.py, VIS-00x.py and specify their path in the config file.

Put the config file under ./Dataset/.round/configs, and run the following command to preprocess and save the dataset.

import round
round.preprocess_dataset(dataset_pth="./Dataset", save_dataset_pth="./Preprocessed_Dataset", config_name="{some config name}")

or for creating dataloaders

import round
dataloader = round.prepare_dataloader(dataset_pth="./Dataset", config_name="{some config name}")