Dataset Viewer

The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.

SEA-CLIP-Tiny — Southeast Asian training blend

The curated Southeast Asian image–text data used to train SEA-CLIP-Tiny (ACCV 2026), packaged as WebDataset shards exactly as they were fed to the training job.

Contents

Prefix Source Pairs
cultural-ground-oe-filt-{id,jv,ms,su,th,vi}/ CulturalGround Open-Ended, filtered 703,475
wit-hf-base-{id,jv,ms,my,th,vi}/ WIT (HuggingFace base split) 487,688
bloom/ Bloom Library 21,534
manifests/ the image/caption manifests the shards were built from —
scripts/ the preparation pipeline that produced the shards —

Per-language counts:

Dataset Ind Jav Msa Sun Mya Tha Vie Total
CG Open-Ended (filtered) 223,098 39,747 161,397 21,238 — 58,433 199,562 703,475
WIT-HF 146,744 11,926 59,504 — 9,941 55,594 203,979 487,688
Bloom 2,169 — 102 — 419 3,009 22 21,534

Not included

The full training mixture also uses two large public corpora, which are not redistributed here — use them from their original sources:

  • CC12M — conceptual_12m (10.97M pairs)
  • SEA-Mammoth — the Southeast Asian subset of MAmmoTH-VL (540K pairs)

Usage

import webdataset as wds

url = "https://huggingface.co/datasets/fassabilf/sea-clip-sea-blend/resolve/main/cultural-ground-oe-filt-id/cultural-ground-oe-filt-id-{000000..000111}.tar"
ds = wds.WebDataset(url).decode("pil").to_tuple("jpg", "txt")

The shard globs for each run are recorded verbatim in the train_data field of params.txt in every model repo.

Licensing

Each source keeps its upstream license — CulturalGround, WIT and Bloom Library respectively. The packaging, filtering and preparation scripts in this repo are MIT. Images are redistributed under the terms of their original sources; if you hold rights to an image here and want it removed, open a discussion on this repo.

Citation

@inproceedings{seacliptiny2026,
  title     = {SEA-CLIP-Tiny: Efficient Multilingual Text-Vision Embedding for Southeast Asian Languages},
  booktitle = {Asian Conference on Computer Vision (ACCV)},
  year      = {2026}
}
Downloads last month
618

Collection including fassabilf/sea-clip-sea-blend