Other
TLV (Touch-Language-Vision)
Touch-language-vision alignment
Paired tactile and scene images selected from VisGel and augmented with sentence-level descriptions through human localization and GPT-4V annotation.
Access links
Primary places to inspect, download, read about, or reproduce the dataset.
Quick facts
- Vehicle type
- Other
- Environment
- Not reported
- Data origin
- Real-world
- Sensor count
- 1
- Ground truth
- Not reported
- Calibration
- Not reported
- Annotations
- Available
Sensor overview
High-level sensor availability before the detailed sensor records below.
Ground truth
Reference scope, method, rate, coverage, and provenance.
Not reported
Calibration and synchronization
Reported calibration level, reproducibility signals, and supporting notes.
- Level
- Not reported
- Processed parameters
- Not reported
- Raw calibration data
- Not reported
- Calibration targets
- Not reported
Calibration data
Calibration reproducibility details are not reported.
Notes
Not reportedKnown or intentional limitations
Reported constraints and characteristics to check before using the dataset.
Descriptions of contact material, texture and hardness are generated from visual images using GPT-4V prompts; the paper does not establish independently measured physical-property ground truth.
All noncontact samples use one uniform sentence, so sentence-level language diversity is concentrated in the contact subset.
Section 3.4 reports 19,834 entries; Table 2 reports 19,843 training examples. Their relationship is unresolved.
Cameras
scene_camera
- Model
- Not reported
- Setup
- Not Reported
- Count
- 1
- Effective cameras
- 1
- Modality
- Not Reported
- Resolution
- Not reported
- Rate
- Not reported
- Shutter
- Not Reported
- Lens type
- Not Reported
- HFOV
- Not reported
- VFOV
- Not reported
- Notes
- VisGel scene images paired with tactile images; TLV selects contact and noncontact frames rather than documenting a full-video release. Camera model, rate and physical count are not given.
IMUs
Not reported
GNSS
Not reported
LiDAR
Not reported
Radar
Not reported
Pitot tube
Not reported
Additional sensors
Other sensor
- Name
- GelSight tactile sensor
- Count
- 1
- Notes
- Tactile images inherited from VisGel; GelSight is the reported sensor label, with no exact model or sampling rate. Represented as a tactile input rather than an extra scene camera.
Annotations
Reported annotation availability and task support.
- Availability
- Available
- Format
- Not reported
- Class count
- Not reported
Semantic segmentation
Not reported
Instance segmentation
Not reported
Object detection
Not reported
Optical flow
Not reported
Depth ground truth
Not reported
Citation
Citation key and BibTeX kept at the end of the page for reference.
@misc{cheng2024tlv,
title={Towards Comprehensive Multimodal Perception: Introducing the Touch-Language-Vision Dataset},
author={Cheng, Ning and Li, You and Gao, Jing and Fang, Bin and Xu, Jinan and Han, Wenjuan},
year={2024},
eprint={2403.09813},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2403.09813v3}
}
Dataset corrections
Notice a specific issue after reviewing the page?
Report an inaccuracy