Chris-davis-thor/Handmovements

Hand Motion & Device Repair Video Dataset

A VideoFolder-format dataset of hand-operation clips covering bicycle drivetrain cleaning, handheld gadget demos, first-person PC maintenance, and appliance / electronics repair with multimeters. Provided by Thordata. Each clip is annotated with a temporal action breakdown, per-segment descriptions, an English caption, and technical metadata.

Layout (HuggingFace VideoFolder)

data/
├── metadata.jsonl          # one row per video; "file_name" links to the clip
└── train/
    ├── clip_01.mp4
    └── ... clip_06.mp4

Load

from datasets import load_dataset
ds = load_dataset("videofolder", "<your-username>/hand-motion-device-repair")
print(ds["train"][0]["caption"])
print(ds["train"][0]["json"]["action_labels"])

Fields

Field	Type	Description
file_name	string	Relative path to the video file
id	string	Clip identifier
provider	string	Data provider (Thordata)
title	string	Short English title
caption	string	One-sentence English description
style	string	Task category (bicycle maintenance / gadget demo / PC maintenance / appliance repair / electronics testing)
setting	string	Scene / environment
device	string	Device or tool being operated
hands	string	How the hands act
action_labels	list<string>	Coarse action tags for classification
segments	list<dict>	Temporal breakdown: start, end (seconds), action, description
duration_s	float	Duration (seconds)
width, height	int	Resolution
fps	float	Frames per second
num_frames_est	int	Estimated frame count

Coverage (6 clips)

Bicycle chain deep-cleaning with brush, handheld electric air-duster demonstration, first-person PC internals dust-blowing, refrigerator evaporator electrical testing, bench electronics component testing, refrigerator coil clamp-meter reading.

Citation

@misc{hand_motion_device_repair,
  title  = {Hand Motion & Device Repair Video Dataset},
  author = {Thordata},
  year   = {2026}
}