PULSE-sample — one complete recording of the PULSE dataset

This repository holds a single complete recording of PULSE, so that the file layout and the data can be inspected without downloading the full release (~85 GB). Every file here is identical to the file at the same path in the full dataset.

• Full dataset: velvet-pine-22/PULSE · DOI 10.57967/hf/11314
• Code: velvet-pine-22/PULSE-code
• Paper: PULSE: A Synchronized Five-Modality Dataset for Sensorimotor Coordination in Long-Horizon Daily Activities, NeurIPS 2026 Evaluations and Datasets Track (OpenReview)

What this sample contains

Recording v1/s1 (volunteer 1, scenario S1 "office desk organization", ~101 s): the five sensor modalities, the two first-person videos, the 28 annotated action segments, and the global metadata tables.

PULSE-sample/
├── data/v1/s1/
│   ├── aligned_emg_100hz.csv        # surface EMG, 8 channels (Myo Armband, right forearm)
│   ├── aligned_eyetrack_100hz.csv   # eye tracking; v1/s1 has right-eye channels only (see the main page)
│   ├── aligned_imu_100hz.csv        # 10 wearable IMUs x 18 columns = 180 columns
│   ├── aligned_mocap_100hz.csv      # 52 Qualisys markers (X, Y, Z, Type) + Frame, Time
│   ├── aligned_pressure_100hz.csv   # finger-level pressure, 25 cells per hand (R1-R25, L1-L25), grams
│   ├── aligned_myo_pose_100hz.csv   # Myo on-board pose label (auxiliary)
│   ├── aligned_myo_quat_100hz.csv   # Myo orientation quaternion (auxiliary)
│   ├── alignment_metadata.json      # per-device start timestamps and alignment window
│   ├── raw/aligned_v1s1_s_Q.tsv     # Qualisys skeleton export (56 segments + 10 fingertip points)
│   └── videos/
│       ├── scene_cam.mp4            # first-person scene camera, 1920x1080, 30 fps
│       └── eye_tracking_overlay.mp4 # same camera with the gaze cursor, at its recorded frame rate
├── annotations/v1/s1.json           # dense segment annotations (primitive / hand / object / text)
├── annotations_flat/segments.csv    # the 28 segments of v1/s1 (rows of the main segments table)
├── metadata/recordings.csv          # manifest of all 337 recordings of the full release
├── metadata/modality_coverage.xlsx  # per-recording modality availability
└── LICENSE                          # CC BY-NC 4.0 with additional use restrictions

The videos have no audio track, and the face of any laboratory member who appears incidentally is pixelated.

How to use this sample

import json
import pandas as pd

ROOT = "data/v1/s1"
emg      = pd.read_csv(f"{ROOT}/aligned_emg_100hz.csv")
eyetrack = pd.read_csv(f"{ROOT}/aligned_eyetrack_100hz.csv")
imu      = pd.read_csv(f"{ROOT}/aligned_imu_100hz.csv")
mocap    = pd.read_csv(f"{ROOT}/aligned_mocap_100hz.csv")
pressure = pd.read_csv(f"{ROOT}/aligned_pressure_100hz.csv")
print([x.shape for x in (emg, eyetrack, imu, mocap, pressure)])

with open("annotations/v1/s1.json") as f:
    ann = json.load(f)
print(len(ann["segments"]), "action segments")

All sensor files share a common 100 Hz timebase (time column, seconds). Each stream is placed on it from its device's recorded start timestamp; EMG is the raw signed Myo signal (about 200 Hz) linearly interpolated to 100 Hz. See Synchronization on the main dataset page for details.

The 28 segments were produced by the same pipeline as the full release: annotation by vision-language models, followed by a human review in which every segment was confirmed or corrected.

Why this recording

Scenario S1 contains a typical mix of grasp / move / place / adjust primitives, and volunteer 1 has all five sensor streams and full-length videos for it. Its eye-tracking file has right-eye channels only, so it is excluded from the eye-tracking benchmarks (listed under Known issues on the main page). The benchmarks in the paper use volunteer-stratified 5-fold cross-validation with the fold files in the code repository; the split column of metadata/recordings.csv is a legacy fixed split and is not the protocol behind the paper's results.

License

Data are released under CC BY-NC 4.0 with additional restrictions on re-identification and surveillance uses; see LICENSE. The companion code is released under MIT in velvet-pine-22/PULSE-code.

Citation

@inproceedings{liang2026pulse,
  title     = {{PULSE}: A Synchronized Five-Modality Dataset for Sensorimotor Coordination in Long-Horizon Daily Activities},
  author    = {Liang, Wolin and Gao, Yang and Yan, Peiyu and Hu, Yuexiang and Xiao, Yingjing and Ying, Xiongfeng and Guo, Yingnian and Zhang, Cheng and Jin, Zhanpeng},
  booktitle = {NeurIPS 2026 Evaluations and Datasets Track},
  year      = {2026},
  url       = {https://openreview.net/forum?id=AoU576r9O2}
}

Dataset DOI: 10.57967/hf/11314. Contact: Yang Gao (gaoyang2025@scut.edu.cn).