OmniFall: From Staged Through Synthetic to Wild, A Unified Multi-Domain Dataset for Robust Fall Detection

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Schneider, David, Marinov, Zdravko, Zhong, Zeyun, Jaus, Alexander, Düger, Rodi, Baur, Rafael, Sarfraz, M. Saquib, Stiefelhagen, Rainer
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866918358670114816
author Schneider, David
Marinov, Zdravko
Zhong, Zeyun
Jaus, Alexander
Düger, Rodi
Baur, Rafael
Sarfraz, M. Saquib
Stiefelhagen, Rainer
author_facet Schneider, David
Marinov, Zdravko
Zhong, Zeyun
Jaus, Alexander
Düger, Rodi
Baur, Rafael
Sarfraz, M. Saquib
Stiefelhagen, Rainer
contents Visual fall detection models trained on small, staged datasets have unclear real-world utility due to limited diversity and inconsistent evaluation protocols. We present OmniFall, a unified benchmark with 80 hours / 15k videos and dense frame-level annotations in a harmonized 16-class taxonomy, spanning three complementary domains: OF-Staged (eight public staged sets, standardized with cross-subject/view splits), OF-Synthetic (12k videos, 17 h; controlled diversity in age, body type, environment, camera), and OF-In-the-Wild (the first test-only benchmark curated from genuine accident videos). OmniFall supports both video classification and timeline segmentation, and its cross-domain protocol isolates staged/synthetic-to-wild generalization. Our results show that carefully designed synthetic data can match or surpass real staged footage on cross-domain transfer, while reducing privacy risk and easing data collection. By combining privacy-amenable synthetic/staged sources with a public, test-only wild target and releasing dense, standardized timelines, OmniFall provides a comprehensive benchmark for privacy-preserving fall detection and fall-related (pre/post-fall) segmentation, enabling robust detectors that generalize to uncontrolled environments. Project page: http://simplexsigil.github.io/omnifall/
format Preprint
id arxiv_https___arxiv_org_abs_2505_19889
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle OmniFall: From Staged Through Synthetic to Wild, A Unified Multi-Domain Dataset for Robust Fall Detection
Schneider, David
Marinov, Zdravko
Zhong, Zeyun
Jaus, Alexander
Düger, Rodi
Baur, Rafael
Sarfraz, M. Saquib
Stiefelhagen, Rainer
Computer Vision and Pattern Recognition
I.2.10; I.5.4
Visual fall detection models trained on small, staged datasets have unclear real-world utility due to limited diversity and inconsistent evaluation protocols. We present OmniFall, a unified benchmark with 80 hours / 15k videos and dense frame-level annotations in a harmonized 16-class taxonomy, spanning three complementary domains: OF-Staged (eight public staged sets, standardized with cross-subject/view splits), OF-Synthetic (12k videos, 17 h; controlled diversity in age, body type, environment, camera), and OF-In-the-Wild (the first test-only benchmark curated from genuine accident videos). OmniFall supports both video classification and timeline segmentation, and its cross-domain protocol isolates staged/synthetic-to-wild generalization. Our results show that carefully designed synthetic data can match or surpass real staged footage on cross-domain transfer, while reducing privacy risk and easing data collection. By combining privacy-amenable synthetic/staged sources with a public, test-only wild target and releasing dense, standardized timelines, OmniFall provides a comprehensive benchmark for privacy-preserving fall detection and fall-related (pre/post-fall) segmentation, enabling robust detectors that generalize to uncontrolled environments. Project page: http://simplexsigil.github.io/omnifall/
title OmniFall: From Staged Through Synthetic to Wild, A Unified Multi-Domain Dataset for Robust Fall Detection
topic Computer Vision and Pattern Recognition
I.2.10; I.5.4
url https://arxiv.org/abs/2505.19889