Ego-Exo Manufacturing
Collection
Ego+exo video of real manufacturing: synchronized video pairs, atomic action labels, skill assessment. For robotics and activity recognition research. • 2 items • Updated
• 1
Request Access to Ego-Exo Manufacturing Synced Pairs
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
This dataset contains temporally synchronized ego+exo video pairs with expert annotations. Please fill in the form below to request access.
Log in or Sign Up to review the conditions and access this dataset content.
Ego-Exo Manufacturing — Synced Pairs
Gated dataset — request access above to download.
This is the Synced Pairs subset of the Ego-Exo Manufacturing dataset. It contains 40 groups of temporally aligned ego+exo video pairs recorded in a real shoe manufacturing facility, paired with atomic action descriptions and expert commentary annotations.
For the standalone (non-synced) sessions, see the public dataset.
Dataset Statistics
| Attribute | Value |
|---|---|
| Synced ego-exo groups | 40 |
| Ego duration | ~36 hours |
| Exo duration | ~36 hours |
| Total video content | ~72 hours |
| Ego frame rate | 30 fps |
| Exo frame rate | ~25–30 fps (varies per group) |
| Resolution | 1080p |
| Format | H.264 / MP4 |
| Audio | No |
| Face privacy | Exo faces Gaussian-blurred |
| Annotations | Atomic action descriptions, expert commentary |
Dataset Structure
ego-exo-manufacturing-synced/
├── takes.json # Root metadata — one entry per group
├── takes/
│ └── groupXX/
│ ├── ego01.mp4 # Trimmed first-person video
│ └── exo01_blurred.mp4 # Trimmed overhead video (faces blurred)
└── annotations/
├── atomic_descriptions/groupXX.json
└── expert_commentary/groupXX.json
Both videos in each group are trimmed so that frame 0 of ego01.mp4 and frame 0 of exo01_blurred.mp4 correspond to the same real-world moment.
Loading the Dataset
import json
from huggingface_hub import hf_hub_download
REPO = "skill-ai/ego-exo-manufacturing-synced"
# Load metadata
takes_path = hf_hub_download(REPO, "takes.json", repo_type="dataset")
with open(takes_path) as f:
takes = json.load(f)
print(f"{len(takes)} groups")
# 40 groups
import cv2
from huggingface_hub import hf_hub_download
REPO = "skill-ai/ego-exo-manufacturing-synced"
GROUP = "group34"
ego_path = hf_hub_download(REPO, f"takes/{GROUP}/ego01.mp4", repo_type="dataset")
exo_path = hf_hub_download(REPO, f"takes/{GROUP}/exo01_blurred.mp4", repo_type="dataset")
ego_cap = cv2.VideoCapture(ego_path)
exo_cap = cv2.VideoCapture(exo_path)
# Frame N from ego_cap == Frame N from exo_cap (same real-world moment)
License
Licensed under Apache 2.0.
Citation
@dataset{egosexomanufacturing2026,
title = {Ego-Exo Manufacturing},
year = {2026},
url = {https://huggingface.co/datasets/skill-ai/ego-exo-manufacturing-synced},
}
- Downloads last month
- 17