laym0nd/Low-Poly-Game-Asset-Images

Low-Poly Game Asset Image Dataset

A synthetic image dataset of low-poly 3D game assets with captions, built for fine-tuning text-to-image models (LoRA / full fine-tune) on the low-poly asset domain.

Structure

dataset/
  images/
    p00001.png          image
    p00001.txt          full prompt (caption)
    p00001.tag.txt      short object-name tag (e.g. "pistol", "tree")
    ...
  examples/
    samples_100.png     preview sheet (100 samples) used in this card
  metadata.jsonl        all fields for every sample
  prompts.jsonl         prompt pool used for generation
  prompts_extra.jsonl   extra prompts for the additional tree samples
  report.txt            validation report (category distribution)

No train/test split is included — split it yourself when training.

Fields

Each line of metadata.jsonl:

Field	Description
id	Sample id, e.g. p00001
category	Object category, e.g. pistol, tree, chest
subject	Object phrase, e.g. a chunky pistol
view	Camera view, e.g. three-quarter front view
background	Background, e.g. plain light gray
tag	Short label: object name only
prompt	Full prompt used to generate the image
file / caption / tag_file	Relative paths to image / prompt / tag

Example record:

{
  "category": "pistol",
  "subject": "a chunky pistol",
  "view": "three-quarter front view",
  "background": "plain light gray",
  "tag": "chunky pistol",
  "id": "p00001",
  "file": "images/p00001.png",
  "caption": "images/p00001.txt",
  "tag_file": "images/p00001.tag.txt",
  "prompt": "a chunky pistol, low-poly 3D game asset, flat-shaded faceted polygons with visible sharp edges, ..."
}

Usage

Build your own split from metadata.jsonl. The file_name + text format used by diffusers, kohya_ss and ai-toolkit is:

{"file_name": "images/p00001.png", "text": "a chunky pistol, low-poly 3D game asset, ..."}

Example — write a 5% validation split:

import json, random

rows = [json.loads(l) for l in open("metadata.jsonl", encoding="utf-8")]
random.Random(0).shuffle(rows)
n_val = int(len(rows) * 0.05)
val, train = rows[:n_val], rows[n_val:]

for name, part in (("train", train), ("val", val)):
    with open(f"{name}.jsonl", "w", encoding="utf-8") as fh:
        for r in part:
            fh.write(json.dumps({"file_name": r["file"], "text": r["prompt"]}) + "\n")

To train with short labels instead of full prompts, replace text with the tag field from metadata.jsonl.

Categories

52 categories in total, including firearms (pistol, rifle, ammo), nature (tree, bush, mushroom, rock, cactus, stump), props (chest, crate, barrel, lantern, potion, book, candlestick), weapons (sword, axe, dagger, bow, spear, shield), buildings (house, hut, tower, well, bridge, windmill), creatures (dragon, owl, cat, frog, rabbit, bird) and more. Exact counts are listed in report.txt.

License

• Dataset: CC0-1.0

The generation code is not included in this folder; it lives in the GitHub repository of this project and is released under the MIT license.