data-archetype/pexels_aesth_2_1024

Pexels Aesthetic 2 1024

This dataset is a bucketed WebDataset-style collection of re-encoded JPEG images with English captions for text-to-image training and evaluation.

The source images are from Pexels and are subject to the Pexels License and Pexels Terms of Service. This dataset is not affiliated with or endorsed by Pexels.

Contents

• Images: 415,366
• Shards: 474 uncompressed TAR files
• Base resolution: 1024
• Format: bucketed_shards_v1
• Images: RGB JPEG, re-encoded at quality 95
• Captions: UTF-8 text, one caption per image
• Metadata: one JSON sidecar per sample

Layout

manifest.json
buckets/<bucket_id>/shard-*.tar

Each TAR shard contains three files per sample key:

<key>.jpg
<key>.txt
<key>.json

The JSON sidecar includes target dimensions, caption provenance fields, JPEG settings, and bucket metadata. The global manifest.json lists buckets, shard paths, sample counts, and export settings.

Buckets

Images are cover-resized and center-cropped into 1024-base aspect buckets with dimensions divisible by 32. The largest buckets are:

bucket	resolution	samples
p1216x832	1216x832	151,350
p832x1216	832x1216	136,972
p832x1152	832x1152	39,987
p1152x832	1152x832	19,574
p1344x768	1344x768	15,010
p768x1344	768x1344	14,356
p896x1088	896x1088	11,430
p896x1152	896x1152	5,533
p1280x768	1280x768	4,878
p1024x1024	1024x1024	3,730

See manifest.json for the full bucket list.

Captions

Captions are selected with a simple waterfall:

1. Model-generated image captions.
2. Human-reviewed fallback captions for a small set of text-heavy images.

Per-sample JSON records which caption variant was used. Captions are descriptive rather than instructional and may still contain occasional mistakes, especially for distant wildlife, fine-grained species, brands, or hard-to-read text.

Loading

Using webdataset:

import webdataset as wds

urls = "buckets/*/shard-*.tar"
dataset = (
    wds.WebDataset(urls)
    .decode("pil")
    .to_tuple("jpg", "txt", "json")
)

for image, caption, meta in dataset:
    ...

Using the Python standard library:

import json
import tarfile
from pathlib import Path

tar_path = next(Path("buckets").rglob("shard-*.tar"))

with tarfile.open(tar_path, "r") as tf:
    for member in tf:
        if not member.isfile() or not member.name.endswith(".txt"):
            continue
        key = member.name[:-4]
        caption = tf.extractfile(member).read().decode("utf-8").strip()
        meta = json.loads(tf.extractfile(tf.getmember(key + ".json")).read())
        image_bytes = tf.extractfile(tf.getmember(key + ".jpg")).read()
        break

Verification

The export was audited after writing:

• Manifest sample count: 415,366
• Shard sample count: 415,366
• Duplicate sample keys: 0
• Missing .jpg / .txt / .json triplets: 0
• Empty captions: 0
• Skipped too small: 28
• Decode errors: 0
• Encode errors: 0