binghaohuang-robot/touch_in_the_wild-dataset

📦 Touch in the Wild

Learning Fine-Grained Manipulation with a Portable Visuo-Tactile Gripper

🔍 Overview

This dataset supports the paper Touch in the Wild: Learning Fine-Grained Manipulation with a Portable Visuo-Tactile Gripper. It contains a wide range of visuo-tactile demonstrations collected using a portable visuo-tactile gripper, spanning both structured lab settings and unstructured real-world environments.

The dataset is divided into three major categories:

1. Four Main Tasks — Core manipulation tasks studied in the paper
2. Indoor Tasks — Additional fine-grained tasks in lab environments
3. In-the-Wild Data — Diverse interactions collected in real-world locations

We also provide a pretraining dataset located in /pretrain_data in .zarr.zip format, containing about 2,700 videos that can be used directly for pretraining experiments.

---

🗂️ Dataset Structure

The dataset is organized into three main parts, each following a consistent folder structure:

1. Four Main Tasks

• Directory: four_tasks/
• Each task has its own subfolder (e.g., test_tube_collection/, fluid_transfer/, etc.)
• Inside each task folder:
  • dataset.zarr.zip – UMI-style dataset
  • mp4/ – folder containing zipped video recordings and metadata

2. Indoor Tasks

• Directory: indoor_data/
• Each indoor task is stored in a separate subfolder
• Inside each task folder:
  • dataset.zarr.zip
  • mp4/

3. In-the-Wild Data

• Directory: in_the_wild/
• Data is grouped by location, not by task
• Inside each location folder:
  • dataset.zarr.zip
  • mp4/

Within the mp4/ folders, videos may be segmented into multiple session folders (e.g., task_sess1/, task_sess2/, etc.). These indicate different data collection sessions, but do not carry additional meaning beyond the session grouping.

📜 Citation

If you use this dataset, please cite:

@article{zhu2025touch,
  title={Touch in the Wild: Learning Fine-Grained Manipulation with a Portable Visuo-Tactile Gripper},
  author={Zhu, Xinyue and Huang, Binghao and Li, Yunzhu},
  journal={arXiv preprint arXiv:2507.15062},
  year={2025}
}