Altogether-FT
(EMNLP 2024) Altogether-FT is an annotated fine-tuning dataset that re-aligns alt-texts into dense captions. It powers altogether captioner to transform Internet-scale quality alt-texts into dense captions, instead of captioning from scratch as naive captions (e.g, "a dog is walking in the park."). It contains 15448 examples for training and 500 examples for evaluation from WIT and DataComp.
@inproceedings{xu2024altogether,
title={Altogether: Image Captioning via Re-aligning Alt-text},
author={Hu Xu, Po-Yao Huang, Xiaoqing Ellen Tan, Ching-Feng Yeh, Jacob Kahn, Christine Jou, Gargi Ghosh, Omer Levy, Luke Zettlemoyer, Wen-tau Yih, Shang-Wen Li, Saining Xie and Christoph Feichtenhofer},
journal={arXiv preprint arXiv:xxxx.xxxxx},
year={2024}
}
Altogether-FT
from datasets import load_dataset
train_dataset = load_dataset("json", data_files="activebus/Altogether-FT/altogether_ft_train.json", field="data")
eval_dataset = load_dataset("json", data_files="activebus/Altogether-FT/altogether_ft_eval.json", field="data")
License
The majority of Altogether-FT is licensed under CC-BY-NC, portions of the project are available under separate license terms: CLIPCap is licensed MIT and open_clip is licensed under the https://github.com/mlfoundations/open_clip license.