Datasets:
A23D 3D Models — Sample for AI/ML Training
A sample from the A23D corpus of 400,000+ human-authored 3D models. This sample is provided for AI teams to inspect the geometry, material quality, and metadata consistency of A23D assets before licensing the full corpus.
Every model is authored in-house to a single production standard — giving consistency across topology, scale, material setup, naming, and metadata that open-source and marketplace asset sets do not provide.
Key highlights
Complete provenance. Every asset is professionally authored and owned by A23D. No web scraping, marketplace aggregation, or AI-generated content.
Single production standard. Every asset follows the same production pipeline, with consistent topology, scale, materials, naming conventions, and metadata across the corpus.
Rich structured metadata. Every model includes hierarchical categories, natural-language captions, product attributes, physical dimensions, and measured geometry statistics.
Component-level semantics. Materials are linked to the geometry they describe, providing structured information about how each object is constructed, not just how it looks.
Physically based materials. Production-quality PBR materials with structured texture data support physically accurate rendering, simulation, and downstream AI workflows.
Enterprise-ready licensing. Commercial licensing, custom dataset curation, and flexible delivery for foundation models, robotics, embodied AI, world models, digital twins, and simulation.
What's in the sample
| Models | 25 |
| Format | GLB (glTF 2.0), single file per model, textures embedded |
| Texture resolution | 4096 × 4096, metallic-roughness workflow |
| Scale | Metres, Y-up, bounding box published per model |
| Preview renders | 25 × 750 × 750 PNG |
| Captions | 25 natural-language descriptions, 93–112 words each |
| Taxonomy | 10 level-2 categories, 21 level-3 |
| Total size | 2.54 GB |
Categories represented: Animals & People, Army & Weapons, Building Elements, Foliage & Landscaping, Furniture, Kitchen & Food, Lighting Fixtures, Structures & Roadside, Toilet, Transport.
Repository structure
metadata.jsonl one row per model, 40 columns — loads directly
metadata/<SKU>.json full record per model, grouped into blocks
models/<SKU>.glb the model — glTF 2.0 binary, textures embedded
previews/<SKU>.png 750 × 750 studio render on a transparent background
Paths in metadata.jsonl are relative to the repository root, and every path is constructible from the SKU alone.
Models in this sample
| SKU | Name | Category | Triangles | Dimensions (m) | Materials | Size (MB) |
|---|---|---|---|---|---|---|
A23DMOD001858A |
Wooden Table Chair Set | Furniture | 105,600 | 1.84 × 0.97 × 1.85 | 4 | 69.8 |
A23DMOD006097A |
Panel Door | Building Elements | 18,186 | 1.01 × 2.09 × 0.17 | 3 | 7.3 |
A23DMOD009439A |
Gazebo | Structures & Roadside | 109,972 | 4.75 × 3.85 × 4.12 | 8 | 217.4 |
A23DMOD012703A |
Sports Car | Transport | 1,764,221 | 1.89 × 1.41 × 4.15 | 28 | 166.4 |
A23DMOD013625A |
Elegant Hanging Light | Lighting Fixtures | 19,600 | 1.06 × 1.93 × 1.00 | 4 | 46.5 |
A23DMOD015025A |
Double Bed | Furniture | 643,421 | 3.55 × 1.15 × 2.55 | 11 | 367.2 |
A23DMOD017743A |
Upholstered Wooden Armchair | Furniture | 610,382 | 2.17 × 0.95 × 1.21 | 4 | 151.2 |
A23DMOD019821A |
Ornamental Houseplants | Foliage & Landscaping | 1,829,416 | 1.04 × 0.90 × 0.85 | 10 | 270.9 |
A23DMOD019825A |
Palm Plant | Foliage & Landscaping | 467,985 | 1.88 × 2.31 × 1.65 | 6 | 143.1 |
A23DMOD020802A |
Wooden Bunk Bed | Furniture | 167,596 | 2.03 × 1.72 × 0.88 | 5 | 105.0 |
A23DMOD037547A |
Apple Rack | Kitchen & Food | 99,108 | 0.45 × 0.61 × 0.45 | 4 | 34.6 |
A23DMOD042822A |
Bicycle | Transport | 133,660 | 0.80 × 1.21 × 2.05 | 8 | 53.0 |
A23DMOD042855A |
Arched Garden Footbridge | Foliage & Landscaping | 104,858 | 2.39 × 0.89 × 0.89 | 2 | 93.0 |
A23DMOD048998A |
Domestic Cat | Animals & People | 63,528 | 0.99 × 0.68 × 0.98 | 1 | 26.0 |
A23DMOD063106A |
Wooden Desk Lamp | Lighting Fixtures | 99,232 | 0.63 × 0.77 × 0.61 | 5 | 31.3 |
A23DMOD064450A |
Washbasin Cabinet | Toilet | 36,800 | 2.06 × 2.07 × 0.52 | 12 | 46.7 |
A23DMOD075483A |
Wooden Windsor Armchair | Furniture | 59,932 | 0.59 × 1.02 × 0.54 | 1 | 30.5 |
A23DMOD088058A |
Upholstered Dining Armchair | Furniture | 105,176 | 0.61 × 0.85 × 0.60 | 3 | 27.3 |
A23DMOD088680A |
Two-Handle Basin Mixer | Toilet | 43,952 | 0.31 × 0.21 × 0.18 | 3 | 37.0 |
A23DMOD116229A |
Writing Desk | Furniture | 19,176 | 1.30 × 0.75 × 0.60 | 3 | 25.0 |
A23DMOD117389A |
Hanging Utensil Rack | Kitchen & Food | 44,616 | 0.76 × 0.48 × 0.10 | 3 | 4.2 |
A23DMOD120766A |
Armoured Fantasy Warrior | Animals & People | 8,999 | 0.81 × 1.72 × 0.47 | 4 | 67.6 |
A23DMOD137661A |
Buttoned Tub Armchair | Furniture | 114,602 | 0.85 × 0.78 × 0.80 | 3 | 71.0 |
A23DMOD196237A |
Scoped Revolver | Army & Weapons | 659,202 | 0.04 × 0.18 × 0.31 | 9 | 158.2 |
A23DMOD208693A |
Village House | Structures & Roadside | 9,796 | 3.86 × 5.77 × 4.15 | 8 | 160.2 |
Metadata
metadata.jsonl loads directly with datasets or pandas. One flat row per model, in this order:
| Field | Type | Description |
|---|---|---|
image |
image | Preview render, decoded by the viewer |
sku |
string | Stable asset ID. Primary key across the full corpus |
name |
string | Short human-readable title |
category_l1 … category_l4 |
string | Category hierarchy, deepest level varies by asset |
attr_material, attr_type, attr_shape, attr_size, attr_back, attr_side, attr_storage, attr_usage |
list | Structured product attributes. Every one is a list, since a model can carry more than one value — empty where the catalogue records nothing |
caption |
string | Natural-language description of the model |
human_authored |
bool | Provenance attestation — true for every asset in the corpus |
format, gltf_version, up_axis, units |
string | Format and coordinate declarations |
width_m, height_m, depth_m |
float | World-space bounding box in metres |
triangles, vertices, polygons, objects, meshes, uv_sets |
int | Measured geometry counts |
has_animations, has_skeleton |
bool | false throughout this sample |
materials_count, pbr_workflow |
int, string | Material count and shading model |
textures_count, texture_resolution, textures_embedded |
int, string, bool | Texture count, resolution and packaging |
total_size_mb |
float | Measured size of the GLB |
model_path |
string | Path to the GLB |
metadata |
string | Path to the full per-model record |
metadata/<SKU>.json carries the same facts grouped rather than flattened — category, attributes, geometry, materials, textures and files — so each block holds one kind of fact. image arrives as file_name in the raw file and is decoded on load.
Usage
from datasets import load_dataset
ds = load_dataset("A23D/a23d-3d-models-sample", split="train")
ds[0]["caption"] # natural-language description
ds[0]["image"] # 750x750 preview render
Text–image pairs for contrastive or generative training:
pairs = [(r["caption"], r["image"]) for r in ds]
The GLB is referenced by path:
import os
import trimesh
from huggingface_hub import snapshot_download
root = snapshot_download("A23D/a23d-3d-models-sample", repo_type="dataset")
scene = trimesh.load(os.path.join(root, ds[0]["model_path"]))
scene.extents # matches width_m / height_m / depth_m
pygltflib, trimesh, Blender, Unreal, Unity, Omniverse and three.js all read these files directly. The Hub's dataset viewer decodes the preview image only; GLB files are downloaded, not previewed inline.
The full corpus
This sample is a small slice of the complete A23D library:
- 400,000+ professionally authored 3D assets
- Broad coverage across architecture, furniture, consumer products, transportation, vegetation, characters, and more
- Consistent topology, metadata, naming conventions, and production standards
- Complete provenance through A23D's in-house production pipeline
- Flexible delivery in customer-required formats and specifications
- Bulk licensing, custom curation, and category-specific collections
Available for licensing for AI and machine learning applications, including foundation models, generative 3D, robotics, embodied AI, world models, digital twins, simulation, synthetic data generation, spatial AI, and other enterprise use cases.
Get in touch:
Website: www.a23d.co/ai
Email: enterprise@a23d.co
License
This sample is released under the A23D Sample Dataset License for evaluation.
You may use it internally to assess the corpus — inspecting, rendering, validating against your pipelines, computing embeddings and benchmarks, and limited-scale fine-tuning to judge suitability. You may not redistribute it, use it commercially, build derivative datasets from it, or deploy or distribute any model trained on it.
Production training rights are granted under a commercial license for the full corpus.
Citation
@misc{a23d_models_sample_dataset_2026,
title = {A23D 3D Models Sample Dataset for AI/ML Training},
author = {A23D},
year = {2026},
publisher = {A2VR Technologies LLP},
url = {https://huggingface.co/datasets/A23D/a23d-3d-models-sample}
}
About A23D
A23D develops enterprise-scale 3D assets and PBR material datasets for artificial intelligence, robotics, simulation and digital content creation — trusted, high-quality, human-authored 3D data for the next generation of AI systems.
One concept, one vision, one structure.
A23D is a brand of A2VR Technologies LLP.
© 2026 A2VR Technologies LLP. All rights reserved.
- Downloads last month
- 12