Anchorworld Dataset

AnchorWorld provides processed training data for an embodied egocentric world simulation framework. The dataset is derived from two public multi-view human-activity datasets, as described in the associated research paper. It was uploaded by user 'lyabc' to Hugging Face and last updated on June 11, 2026.

Use Cases

• Train models for egocentric world simulation based on 3D human motion data.
• Customize spatial simulations using the spatially grounded anchor views mentioned in the description.
• Benchmark performance of frameworks for human activity understanding from multi-view data.

Strengths

• Derived from two public multi-view human-activity datasets, suggesting a foundation in established data.
• Associated with a published research paper and project page, indicating academic rigor.

Limitations

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

Provenance

Source

Derived from two public multi-view human-activity datasets.

Collection Method

Processed for training the AnchorWorld framework.

Freshness

Last updated 2026-06-11 06:03:43; freshness should be verified.