Objaverse-XL: Over 10 Million 3D Objects for AI Training

Over 10 million 3D objects form this open dataset, which is more than an order of magnitude larger than its predecessor. AllenAI released Objaverse-XL in 2023 to train the Zero123-XL foundation model for 3D tasks. The dataset is described as being much more diverse than the earlier 800K-object Objaverse 1.0.

Use Cases

• Training 3D foundation models based on the scale of over 10 million objects.
• Improving 3D generalization abilities for AI systems based on the dataset's stated purpose.
• Conducting research in computer graphics and synthetic data generation based on the dataset's content.
• Benchmarking 3D object recognition or generation algorithms based on the large, diverse collection.

Strengths

• Contains over 10 million 3D objects, a large scale for training.
• Described as being more diverse than the previous 800K-object version.
• Used to train a named foundation model, Zero123-XL, suggesting practical utility.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

allenai

Freshness

Last updated 2023-10-31 16:46:54; freshness should be verified.