data-archetype/pexels_aesth_bucketed_512

A captioned and aspect ratio bucketed dataset of about 200k pexel photos with subjectively good aesthetic quality, with a diverse range of themes

This version is scaled to approx 512^2 resolution. See below for more details

Bucketed Shards Dataset

This repository contains a bucketed-shards export (uncompressed TAR shards).

Format

• Format: bucketed_shards_v1
• Created: 2026-01-08T09:51:12.478529+00:00
• Export ID: export-2026-01-08T09:51:12.478529+00:00
• Manifest: manifest.json

Directory layout:

• manifest.json (global metadata + per-bucket shard listing)
• buckets/<bucket_id>/shard-*.tar

Each TAR shard contains 3 files per sample:

• <key>.jpg (RGB JPEG)
• <key>.txt (caption text, UTF-8, newline-terminated)
• <key>.json (per-sample metadata: resize/crop/jpeg params + source size/path)

Image preprocessing

All images are processed deterministically per-sample:

• EXIF transpose, convert to RGB
• Cover-resize using torch CPU bicubic interpolation with antialiasing (mode=bicubic, antialias=True)
• Never upsample: samples that would require upscaling are skipped (too_small_policy=drop)
• Crop to the bucket target size (crop_strategy=corner, allowed corners [2, 3])

Resize/crop details:

• Cover scale is scale = max(target_w / src_w, target_h / src_h); if scale > 1, the sample is skipped.
• After resize, a crop box is chosen deterministically from the sample key (sha256 of image_id).
  • Corner strategy chooses a corner from allowed_corners where 0=TL, 1=TR, 2=BL, 3=BR (optional small jitter for corner_jitter).

JPEG encoding:

• quality 95
• subsampling policy adaptive_scale (adaptive threshold 0.85)

Buckets / resolutions

• Buckets follow SDXL-style proto buckets defined at a 1024×1024 base.
• Base resolution(s): [512]
• In single-res exports, bucket_id is the proto (1024-base) bucket, e.g. p1024x1024.
• In multi-res exports, buckets are namespaced by base resolution: r<base>_<proto>, e.g. r512_p1024x1024.
• The actual target resolution for each bucket (scaled by the per-bucket base resolution and divisible=32) is stored in:
  • manifest.json → buckets[<bucket_id>].scaled.w/h (and base_resolution)
  • each sample’s <key>.json → base.w/h, plus resize and crop details

Bucket IDs (preview): p1024x1024, p1024x960, p1088x896, p1088x960, p1152x832, p1152x896, p1216x832, p1280x768, p1344x704, p1344x768, p1408x704, p1472x704, p1536x640, p1600x640, p1664x576, p1728x576, p1792x576, p1856x512, p1984x512, p2048x512, … (+13 more)

Bucket distribution:

bucket_id	target_w×h	aspect	count
p832x1216	416×608	0.684	89,115
p1216x832	608×416	1.462	73,621
p832x1152	416×576	0.722	16,357
p896x1152	448×576	0.778	8,652
p1152x832	576×416	1.385	7,421
p1344x768	672×384	1.750	5,771
p768x1344	384×672	0.571	4,673
p1280x768	640×384	1.667	2,534
p1024x1024	512×512	1.000	1,912
p1152x896	576×448	1.286	1,647
p896x1088	448×544	0.824	1,485
p768x1280	384×640	0.600	1,244
p1344x704	672×352	1.909	402
p960x1088	480×544	0.882	388
p1088x896	544×448	1.214	381
p960x1024	480×512	0.938	300
p1024x960	512×480	1.067	246
p1088x960	544×480	1.133	236
p1408x704	704×352	2.000	145
p1472x704	736×352	2.091	133
p1536x640	768×320	2.400	69
p704x1408	352×704	0.500	69
p704x1472	352×736	0.478	35
p1600x640	800×320	2.500	34
p1664x576	832×288	2.889	25
p1728x576	864×288	3.000	19
p1792x576	896×288	3.111	15
p1856x512	928×256	3.625	8
p2048x512	1024×256	4.000	8
p640x1536	320×768	0.417	6
p1984x512	992×256	3.875	1
p512x2048	256×1024	0.250	1
p640x1600	320×800	0.400	1

Caption selection (waterfall)

Captions are selected from dataset.sqlite using the first matching selector (highest priority wins). Within the same selector, the newest caption source is preferred.

Caption provenance:

• Per-sample <key>.json includes caption_source_id (int, from dataset.sqlite).
• manifest.json includes a caption_sources table mapping caption_source_id → backend/model/created_at plus prompt hashes (not prompt text).

Caption priority (waterfall) + planned usage:

#	variant	backend	model	planned_images
0	caption_gemini			34,898
1	caption_ministral_14b_2512			182,056

Available caption variants (top 30):

selected	variant	images_with_ok_caption
✓	caption_ministral_14b_2512	182,056
✓	caption_gemini	34,898

Missing caption policy: drop

Export summary

• images_seen: 216,954
• images_exported: 216,954
• skipped_no_caption: 0
• skipped_too_small: 0
• decode_errors: 0
• encode_errors: 0

Efficient loading

Recommended

Treat this as a webdataset-style collection of tar shards:

• Prefer sequential reads of tar files for throughput.
• Shuffle at the shard level (and optionally within-shard) for good randomness without expensive random I/O.
• Use manifest.json to list buckets and shards.

Python (webdataset)

import webdataset as wds

urls = "buckets/*/shard-*.tar"  # glob; adjust if you want a single bucket only
ds = (
    wds.WebDataset(urls)
    .decode("pil")            # decodes .jpg to PIL.Image
    .to_tuple("jpg", "txt", "json")
)
for jpg, caption, meta in ds:
    ...

Python (tarfile, no extra deps)

import io, json, tarfile
from pathlib import Path

tar_path = next(Path("buckets").rglob("shard-*.tar"))
with tarfile.open(tar_path, "r") as tf:
    members = tf.getmembers()
    for m in members:
        if not m.name.endswith(".txt"):
            continue
        key = m.name[:-4]
        caption = tf.extractfile(m).read().decode("utf-8").strip()
        meta = json.loads(tf.extractfile(tf.getmember(key + ".json")).read().decode("utf-8"))
        jpg_bytes = tf.extractfile(tf.getmember(key + ".jpg")).read()
        ...