Skip to content

Feature Request: Add TLabel format as a supported conversion target #39

Description

@liesliy

Context

Forge is a great toolkit for converting between robotics dataset formats (RLDS, LeRobot, etc.). As tactile sensing becomes increasingly important in robot learning, there's a growing need for format conversion that includes tactile data.

Proposal

Add TLabel as a supported format for tactile sensor data. TLabel provides:

  • Unified schema for tactile data across 10+ sensor types (GelSight, PaXini, Daimon, UniVTAC, VTouch, YCB-Slide, TacQuad, etc.)
  • Dataset adapters for loading existing tactile datasets into standardized format
  • Real-time sensor adapters for live data capture (PaXini GEN3 SDK, Daimon DM-Tac USB)
  • PyPI package: pip install tlabel (MIT license, 14K+ downloads)

Use case

A researcher has LeRobot dataset with visual + joint data, and wants to augment it with tactile data from GelSight sensors. Forge + TLabel could provide the bridge: convert LeRobot → TLabel (adding tactile channels) → back to LeRobot with enriched modalities.

This would make Forge the go-to tool for multi-modal robotics data format conversion, including the tactile modality.

Refs:

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions