UnoBench: Target-Centric Obstruction Reasoning for Robotic Grasping

UnoBench is a benchmark for target-centric obstruction reasoning in robotic grasping under cluttered scenes. It is built upon MetaGraspNetV2 and extends the initial idea of FreeGraspData. The dataset was authored by FBK-TeV and was last updated on the Hugging Face platform in May 2026.

Use Cases

• Benchmarking robotic grasping algorithms based on obstruction reasoning.
• Training models to identify blocking objects in cluttered scenes.
• Evaluating methods for target-centric manipulation planning.

Strengths

• Built upon the established MetaGraspNetV2 framework.
• Extends the initial concept of FreeGraspData.
• Associated code and checkpoints are available via a linked GitHub repository.

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

FBK-TeV

Collection Method

Built upon MetaGraspNetV2 and extends FreeGraspData.

Freshness

Last updated 2026-05-31 23:50:51; freshness should be verified.