TLabel: Cross-Sensor Tactile Manipulation Data with Semantic Annotation

TLabel Format is a unified annotation framework for tactile manipulation data. It includes validation data from two sensors: Daimon-Infinity (94 episodes, 6 tasks) and PaXini PXCap (15 episodes, 7 tasks). The dataset was published by Xi Luo on figshare in June 2026.

Use Cases

• Benchmarking tactile annotation systems based on the declared ten semantic dimensions.
• Training models for slip detection based on the reported +10.35% F1 improvement.
• Evaluating cross-sensor generalization performance based on the +7.93% accuracy metric.
• Comparing vision-based and Hall-effect tactile sensing paradigms based on the complementary results.

Strengths

• Validated on two sensors with fundamentally different physics: Daimon-Infinity and PaXini PXCap.
• Achieved zero hard errors across 590K+ observations.
• Demonstrated measurable downstream benefits in generalization accuracy and slip-risk detection.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is small (1.9 MB), which may constrain the scope of analysis.

Provenance

Source

figshare

Collection Method

Data collected and annotated using the TLabel Format schema.

Freshness

Last updated 2026-06-01 01:24:17; freshness should be verified.