README.md

metadata

license: mit
task_categories:
  - zero-shot-classification
  - image-to-text
language:
  - en
pretty_name: yfcc15m
size_categories:
  - 10M<n<100M

YFCC15M subset used for VLMs

This dataset contains the ~15M subset of YFCC100M used for training the models in the paper Quality Not Quantity: On the Interaction between Dataset Design and Robustness of CLIP. The metadata provided in this repo contains both the page-urls and image-download-urls for downloading the dataset. This dataset can be easily downloaded with img2dataset:

img2dataset --url_list yfcc15m_final_split_pageandimageurls.csv --input_format "csv" --output_format webdataset --output_folder images --processes_count 2 --thread_count 8 --resize_mode no --enable_wandb True