For EnterpriseAI / ML / Gen-AI training datasets

Ground-truth 3D training data for AI, ML and generative models

We offer the largest proprietary library of human-authored 3D assets for AI and machine learning. Created entirely by our in-house team, every 3D model is paired with our PBR materials, providing structured ground truth at both the geometry and material level. Each asset includes metadata and semantic annotations at object and component level, providing fine-grained training signals for AI.

577,000+

3D Assets

Consistent scale, orientation, topology, pivots, materials, metadata, and labels across the entire dataset.

100%

Human-authored

Every asset is created by our in-house artists, with no AI-generated geometry or synthetic source content.

30+

Structured fields per asset

Rich object- and component-level metadata designed for filtering, labeling, and supervised AI model training.

625+

Categories

Standardized categories covering 3D Models, PBR Textures, Decals, Foliage, Characters, and IES Profiles.

Unlike traditional 3D datasets assembled from multiple sources or marketplace collections aggregated from many independent creators, every asset in our library is created through a standardized in-house pipeline, ensuring consistent geometry, materials, metadata, and semantic annotations — with clear, uniform provenance across the entire corpus for reliable training and evaluation across large-scale AI workflows.

Tell us about your project

Our enterprise team will get in touch with you.

Prefer email? Reach us at enterprise@a23d.co

3D Training Dataset for AI and Machine Learning

A23D provides one of the world's largest proprietary datasets of human-authored 3D assets for artificial intelligence and machine learning. The dataset contains more than 577,000 assets, including 3D models, PBR materials, decals, foliage, characters, and IES lighting profiles spanning over 625 categories. Every asset is created entirely by A23D's in-house team using a standardized production pipeline.

Structured Geometry and Material Ground Truth

Every 3D model is paired with matching PBR materials, providing structured ground truth at both the geometry and material level. Material data includes physically based rendering properties such as albedo, normal, roughness, metallic, opacity, and subsurface information. The relationship between geometry and materials is preserved across the entire dataset.

Metadata and Semantic Annotations

Each asset includes structured metadata and semantic annotations at both the object and component level. Metadata can include categories, materials, dimensions, hierarchy information, semantic labels, and component relationships. These annotations provide rich training signals for supervised learning, scene understanding, semantic segmentation, and multimodal AI systems.

Built for Generative 3D, Robotics, and World Models

The dataset is designed for generative 3D models, text-to-3D systems, image-to-3D systems, world models, robotics simulation, embodied AI, synthetic data generation, computer vision, neural rendering, and multimodal foundation model training. Standardized scale, orientation, topology, pivots, materials, metadata, and semantic labels enable reliable large-scale training and evaluation workflows.

Human-Authored 3D Data

All assets are human-authored and created entirely by A23D's in-house artists. No AI-generated geometry or synthetic source assets are included in the dataset. This provides consistent quality, traceability, and clear dataset provenance across the entire corpus.

Enterprise Dataset Licensing and Delivery

A23D offers enterprise licensing for AI training, research, model development, and production deployments. Datasets can be delivered through bulk exports, APIs, cloud storage integrations, and reproducible dataset manifests suitable for enterprise machine learning workflows.