# rom1504/img2dataset

Easily turn large sets of image urls to an image dataset. Can download, resize and package 100M urls in 20h on one machine.

Repository: https://github.com/rom1504/img2dataset
Canonical: https://ross.abutalabs.com/products/img2dataset
Language: Python
License: MIT
License Family: permissive
Topics: deep-learning, dataset, big-data, image, multimodal, image-dataset, download-images
Last push: 2025-10-19T21:22:50+00:00

## Health v2 (maintenance only)
Score: 56/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 47, release rhythm 42, longevity 100
- inputs: {"age_days": 1848, "days_push": 318, "days_rel": 389, "gap_med": 0, "n_releases_24m": 2}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 4443, forks 376 (observed 2026-08-28T04:08:50.346193+00:00)

## What it is
A Python tool that downloads large sets of image URLs and packages them into machine learning datasets, with resizing and caption support. It can process 100M URLs in 20 hours on a single machine and powers datasets like LAION-5B.

## Use cases
- download images from a list of urls into a dataset
- build an image-text dataset for training models
- download and resize millions of images quickly
- create a laion-style dataset from image url list
- package image urls into webdataset shards
- download mscoco or cc3m image datasets

## When to choose
- you need to turn a large url list into a usable image dataset at scale
- you want image+caption pairs for training multimodal models
- you need fast, parallel, resumable image downloading with resizing

## When to avoid
- you only need a handful of images and a simple downloader suffices
- you need to scrape and discover image urls rather than download from an existing list
- you need dataset labeling or annotation, not downloading

## Facets
- artifact type: cli-tool
- maturity: active
- function: image-processing, etl, data-generation
- domain: machine-learning, computer-vision, big-data
- platform: python, cli, windows
- tags: image-dataset, webdataset, image-download, multimodal, url-to-dataset, resizer, data-engineering, linux, macos

## Member repositories
- rom1504/img2dataset (main) score 56

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:50.346193+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:20:49.144601+00:00, confidence not recorded.
  - readme: https://github.com/rom1504/img2dataset (fetched 2026-08-28T04:08:50.346193+00:00, sha 045b437a77be)
  - registry_pypi: https://pypi.org/pypi/img2dataset/json (fetched 2026-08-29T09:07:51.766729+00:00, sha cfb759400670)
- Data as of 2026-08-30T08:39:29.467469+00:00.