Mobjaverse: 19,000 Rigged 3D Models with Skeletal Animations

Mobjaverse is a large-scale dataset of rigged 3D models derived from Objaverse-XL, curated for research on skeletal animation and motion generation. It contains approximately 19,000 models spanning around 5,000 distinct skeletal topologies and includes about 2 million motion frames. The dataset was created by author 'duckduckplz' and was last updated on June 12, 2026.

Use Cases

• Training models for skeletal animation understanding based on rigged 3D models.
• Developing motion generation algorithms based on the dataset's ~2 million motion frames.
• Conducting articulated 3D shape analysis based on models with ~5k distinct skeletal topologies.
• Creating priors for monocular video-to-animation tasks, as referenced in the associated TopoCap paper.

Strengths

• Contains approximately 19,000 rigged 3D models, providing a large-scale resource.
• Encompasses about 5,000 distinct skeletal topologies, offering diversity in model structure.
• Includes roughly 2 million motion frames for animation research.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset's license is unknown, which may restrict usage.
• Row count and file formats are unknown, which may limit suitability assessment.

Provenance

Source

Curated from the Objaverse-XL dataset.

Collection Method

Derived and curated as described in the paper 'TopoCap: Learning Topology-Agnostic Motion Priors for Monocular Video-to-Animation'.

Freshness

Last updated 2026-06-12 02:03:20; freshness should be verified.