Pair Touch 13m
Multi-modal tactile dataset with pose, force, and tactile sensor data. | Config | Description | Sensors | |--------|-------------|---------| | | Pose estimation data | tac02/xela + camera | | | Force measuremen...
| Entity Passport | |
| Registry ID | hf-dataset--borisguo--pair_touch_13m |
| Provider | huggingface |
Cite this dataset
Academic & Research Attribution
@misc{hf_dataset__borisguo__pair_touch_13m,
author = {BorisGuo},
title = {Pair Touch 13m Dataset},
year = {2026},
howpublished = {\url{https://huggingface.co/datasets/BorisGuo/pair_touch_13m}},
note = {Accessed via Free2AITools Knowledge Fortress}
} π¬Technical Deep Dive
Full Specifications [+]βΎ
π¬ Why this score?
The Nexus Index for Pair Touch 13m aggregates Popularity (P:0), Velocity (V:0), and Credibility (C:0). The Utility score (U:0) represents deployment readiness, context efficiency, and structural reliability within the Nexus ecosystem.
π Source Links (Click to verify)
ποΈ Data Preview
Row-level preview not available for this dataset.
Schema structure is shown in the Field Logic panel when available.
π Explore Full Dataset β𧬠Field Logic
Schema not yet indexed for this dataset.
Dataset Specification
configs:
- config_name: pose_data
data_files: "pose_data/metadata.jsonl" - config_name: force_data
data_files: "force_data/metadata.jsonl" - config_name: tacniq_gsmini
data_files: "tacniq_gsmini/metadata.jsonl" - config_name: xela_9dtact
data_files: "xela_9dtact/metadata.jsonl"
PairTouch 13M Dataset
Multi-modal tactile dataset with pose, force, and tactile sensor data.
Configs
| Config | Description | Sensors |
|---|---|---|
pose_data |
Pose estimation data | tac02/xela + camera |
force_data |
Force measurement data | tac02/xela + gelsight |
tacniq_gsmini |
TacNIQ + GSMini data | tacniq + gsmini |
xela_9dtact |
XELA + 9DTact data | xela + 9dtact |
Usage
from datasets import load_dataset
Load specific config
ds = load_dataset("BorisGuo/pair_touch_13m", "pose_data")
Tactile Sensors
tac02 / tacniq
- Taxel layout: 11 x 6 = 66 taxels
- Force type: Z-axis normal force only
- Dead taxels: Index 0 (top-left) and 5 (top-right) always read 0, should be ignored
Taxel indices:
[ 0][ 1][ 2][ 3][ 4][ 5] <- Row 0 (index 0, 5 are dead)
[ 6][ 7][ 8][ 9][10][11] <- Row 1
...
[60][61][62][63][64][65] <- Row 10
xela
- Taxel layout: 4 x 6 = 24 taxels
- Force type: 3D force (X, Y, Z axes), total 72 values (24 x 3)
- Data format: Flattened as [x0, y0, z0, x1, y1, z1, ..., x23, y23, z23]
Sampling Frequency
force_data (1 tactile sample per frame)
| Dataset | Frame Rate | Tactile Sensor |
|---|---|---|
| 9dtact_force_h5 | ~10 Hz | - |
| gelsight_force_h5 | ~25 Hz | - |
| tac02_force_h5 | ~200 Hz | tac02 |
| xela_force_h5 | ~100 Hz | xela |
pose_data (multiple tactile samples per frame)
| Dataset | Frame Rate | Samples/Frame | Effective Tactile Rate |
|---|---|---|---|
| tac02_pose_h5 | ~7.5 Hz | 20 | ~150 Hz |
| xela_pose_h5 | ~7.5 Hz | 10 | ~75 Hz |
| gelsight_pose_h5 | ~7.5 Hz | - | - |
| 9dtact_pose_h5 | ~7.5 Hz | - | - |
Paired sensor data
| Dataset | Frame Rate | Sensors |
|---|---|---|
| tacniq_gsmini | 10 Hz | tacniq + gsmini |
| xela_9dtact | 10 Hz | xela + 9dtact |
Data Fields
file_name: Path to image fileepisode_id: Episode identifierframe_idx: Frame index within episodetimestamp: Timestamprotation: 3D rotationtranslation: 3D translationtactile: tac02 sensor matrix (11x6 taxels, Z-axis only)xela: XELA sensor matrix (4x6x3 taxels, 3D force)tacniq: TacNIQ sensor matrix (same as tac02)video: Episode video path
Data Dimensions
images: (N, S, 224, 224, 3) - N frames, S samples per frametactile(pose_data): (N, 20, 66) - N frames, 20Hz sampling, 66 taxels (11x6)xela(pose_data): (N, 10, 72) - N frames, 10Hz sampling, 72 values (4x6x3)tactile/xela(force_data): (N, 66) or (N, 72) - 1 sample per frame
Preprocessing
Use preprocess.py for data processing:
# Parse H5 files to JSON/images
python preprocess.py extract
Check H5 structure (without parsing)
python preprocess.py extract --check
Update metadata (add heatmap/video paths)
python preprocess.py extract --update
Generate heatmaps
python preprocess.py heatmap
python preprocess.py heatmap --test # Test mode
python preprocess.py heatmap --type tac02_pose # Specific type
Generate xela marker flow (point displacement visualization)
python preprocess.py marker_flow
python preprocess.py marker_flow --test
python preprocess.py marker_flow --type xela_pose
Generate videos
python preprocess.py video
python preprocess.py video --test
Pack images into tar files (reduce file count for HF upload)
python preprocess.py pack # Pack only
python preprocess.py pack --delete # Pack and delete original PNGs
Unpack images from tar files
python preprocess.py unpack # Unpack only
python preprocess.py unpack --delete # Unpack and delete tar files
Clean up (delete all PNGs, keep only videos)
python preprocess.py clean
Upload to Hugging Face
python preprocess.py upload # εͺδΈδΌ /ζ΄ζ°οΌδΈε ι€θΏη«―ζδ»Ά
python preprocess.py upload --sync # εζ₯樑εΌοΌε ι€θΏη«―εε¨δ½ζ¬ε°δΈεε¨ηζδ»Ά
Full pipeline (extract -> heatmap -> video -> update)
python preprocess.py all
Heatmap Types
| Type | Description |
|---|---|
tac02_pose |
pose_data/tac02_pose_h5 |
xela_pose |
pose_data/xela_pose_h5 |
pose |
All pose_data |
tac02_force |
force_data/tac02_force_h5 |
xela_force |
force_data/xela_force_h5 |
force |
All force_data |
tacniq_gsmini |
tacniq_gsmini |
xela_9dtact |
xela_9dtact |
all |
Everything |
Video Types
| Type | Description |
|---|---|
9dtact_force |
force_data/9dtact_force_h5 |
xela_force |
force_data/xela_force_h5 |
gelsight_force |
force_data/gelsight_force_h5 |
tac02_force |
force_data/tac02_force_h5 |
gelsight_pose |
pose_data/gelsight_pose_h5 |
9dtact_pose |
pose_data/9dtact_pose_h5 |
tac02_pose |
pose_data/tac02_pose_h5 |
xela_pose |
pose_data/xela_pose_h5 |
xela_pose_flow |
pose_data/xela_pose_h5 marker_flow (video_flow.mp4) |
xela_force_flow |
force_data/xela_force_h5 marker_flow (video_flow.mp4) |
xela_9dtact_flow |
xela_9dtact marker_flow (video_flow.mp4) |
tacniq_gsmini |
tacniq_gsmini (video_gsmini.mp4 + video_tacniq.mp4) |
xela_9dtact |
xela_9dtact (video_9dtact.mp4 + video_xela.mp4) |
all |
Everything |
Social Proof
AI Summary: Based on Hugging Face metadata. Not a recommendation.
π‘οΈ Dataset Transparency Report
Verified data manifest for traceability and transparency.
π Identity & Source
- id
- hf-dataset--borisguo--pair_touch_13m
- source
- huggingface
- author
- BorisGuo
- tags
- modality:imageregion:us
βοΈ Technical Specs
- architecture
- null
- params billions
- 0.013
- context length
- null
π Engagement & Metrics
- likes
- 2
- downloads
- 89,218
Free2AITools Constitutional Data Pipeline: Curated disclosure mode active. (V15.x Standard)