Self-supervised Traffic Accident Detection by Motion-Conditioned Diffusive Frame Prediction

Xinyan Cui; Lei Lei Li; Yachuang Chai; Jianwu Fang; Wushour Silamu

doi:10.1007/978-981-97-1087-4_28

Self-supervised Traffic Accident Detection by Motion-Conditioned Diffusive Frame Prediction

Xinyan Cui
, Lei Lei Li
, Yachuang Chai
, Jianwu Fang
, Wushour Silamu

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

1 Scopus citations

Abstract

The detection of anomalies using dashcam videos is crucial for autonomous driving or driver assistance systems. The scarcity of diverse accident videos and the complex environment variations during driving significant struggle for accident detection. Leveraging the advancements in frame prediction-based accident detection methods and diffusion models, we propose a frame prediction framework based on motion-conditioned diffusion (MCD-TAD). This framework combines optical flow features with appearance features in consecutive video frames using a latent diffusion model to better capture and utilize spatiotemporal cues. Extensive evaluations on two large-scale accident datasets, namely AnAn Accident Detection (A3D) dataset and DADA-2000 dataset, validate the effectiveness of the MCD-TAD for traffic accident detection in dashcam videos.

Original language	English
Title of host publication	Proceedings of 3rd 2023 International Conference on Autonomous Unmanned Systems (3rd ICAUS 2023) - Volume III
Editors	Yi Qu, Mancang Gu, Yifeng Niu, Wenxing Fu
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	292-303
Number of pages	12
ISBN (Print)	9789819710867
DOIs	https://doi.org/10.1007/978-981-97-1087-4_28
State	Published - 2024
Externally published	Yes
Event	3rd International Conference on Autonomous Unmanned Systems, ICAUS 2023 - Nanjing, China Duration: 9 Sep 2023 → 11 Sep 2023

Publication series

Name	Lecture Notes in Electrical Engineering
Volume	1173 LNEE
ISSN (Print)	1876-1100
ISSN (Electronic)	1876-1119

Conference

Conference	3rd International Conference on Autonomous Unmanned Systems, ICAUS 2023
Country/Territory	China
City	Nanjing
Period	9/09/23 → 11/09/23

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being

Keywords

Traffic accident detection
frame prediction
latent diffusion model

Access to Document

10.1007/978-981-97-1087-4_28

Cite this

Cui, X., Li, L. L., Chai, Y., Fang, J., & Silamu, W. (2024). Self-supervised Traffic Accident Detection by Motion-Conditioned Diffusive Frame Prediction. In Y. Qu, M. Gu, Y. Niu, & W. Fu (Eds.), Proceedings of 3rd 2023 International Conference on Autonomous Unmanned Systems (3rd ICAUS 2023) - Volume III (pp. 292-303). (Lecture Notes in Electrical Engineering; Vol. 1173 LNEE). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-981-97-1087-4_28

Cui, Xinyan ; Li, Lei Lei ; Chai, Yachuang et al. / Self-supervised Traffic Accident Detection by Motion-Conditioned Diffusive Frame Prediction. Proceedings of 3rd 2023 International Conference on Autonomous Unmanned Systems (3rd ICAUS 2023) - Volume III. editor / Yi Qu ; Mancang Gu ; Yifeng Niu ; Wenxing Fu. Springer Science and Business Media Deutschland GmbH, 2024. pp. 292-303 (Lecture Notes in Electrical Engineering).

@inproceedings{b2bd399aeebe42a4a38efef557560005,

title = "Self-supervised Traffic Accident Detection by Motion-Conditioned Diffusive Frame Prediction",

abstract = "The detection of anomalies using dashcam videos is crucial for autonomous driving or driver assistance systems. The scarcity of diverse accident videos and the complex environment variations during driving significant struggle for accident detection. Leveraging the advancements in frame prediction-based accident detection methods and diffusion models, we propose a frame prediction framework based on motion-conditioned diffusion (MCD-TAD). This framework combines optical flow features with appearance features in consecutive video frames using a latent diffusion model to better capture and utilize spatiotemporal cues. Extensive evaluations on two large-scale accident datasets, namely AnAn Accident Detection (A3D) dataset and DADA-2000 dataset, validate the effectiveness of the MCD-TAD for traffic accident detection in dashcam videos.",

keywords = "Traffic accident detection, frame prediction, latent diffusion model",

author = "Xinyan Cui and Li, \{Lei Lei\} and Yachuang Chai and Jianwu Fang and Wushour Silamu",

note = "Publisher Copyright: {\textcopyright} Beijing HIWING Scientific and Technological Information Institute 2024.; 3rd International Conference on Autonomous Unmanned Systems, ICAUS 2023 ; Conference date: 09-09-2023 Through 11-09-2023",

year = "2024",

doi = "10.1007/978-981-97-1087-4\_28",

language = "英语",

isbn = "9789819710867",

series = "Lecture Notes in Electrical Engineering",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "292--303",

editor = "Yi Qu and Mancang Gu and Yifeng Niu and Wenxing Fu",

booktitle = "Proceedings of 3rd 2023 International Conference on Autonomous Unmanned Systems (3rd ICAUS 2023) - Volume III",

}

Cui, X, Li, LL, Chai, Y, Fang, J & Silamu, W 2024, Self-supervised Traffic Accident Detection by Motion-Conditioned Diffusive Frame Prediction. in Y Qu, M Gu, Y Niu & W Fu (eds), Proceedings of 3rd 2023 International Conference on Autonomous Unmanned Systems (3rd ICAUS 2023) - Volume III. Lecture Notes in Electrical Engineering, vol. 1173 LNEE, Springer Science and Business Media Deutschland GmbH, pp. 292-303, 3rd International Conference on Autonomous Unmanned Systems, ICAUS 2023, Nanjing, China, 9/09/23. https://doi.org/10.1007/978-981-97-1087-4_28

Self-supervised Traffic Accident Detection by Motion-Conditioned Diffusive Frame Prediction. / Cui, Xinyan; Li, Lei Lei; Chai, Yachuang et al.
Proceedings of 3rd 2023 International Conference on Autonomous Unmanned Systems (3rd ICAUS 2023) - Volume III. ed. / Yi Qu; Mancang Gu; Yifeng Niu; Wenxing Fu. Springer Science and Business Media Deutschland GmbH, 2024. p. 292-303 (Lecture Notes in Electrical Engineering; Vol. 1173 LNEE).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Self-supervised Traffic Accident Detection by Motion-Conditioned Diffusive Frame Prediction

AU - Cui, Xinyan

AU - Li, Lei Lei

AU - Chai, Yachuang

AU - Fang, Jianwu

AU - Silamu, Wushour

N1 - Publisher Copyright: © Beijing HIWING Scientific and Technological Information Institute 2024.

PY - 2024

Y1 - 2024

N2 - The detection of anomalies using dashcam videos is crucial for autonomous driving or driver assistance systems. The scarcity of diverse accident videos and the complex environment variations during driving significant struggle for accident detection. Leveraging the advancements in frame prediction-based accident detection methods and diffusion models, we propose a frame prediction framework based on motion-conditioned diffusion (MCD-TAD). This framework combines optical flow features with appearance features in consecutive video frames using a latent diffusion model to better capture and utilize spatiotemporal cues. Extensive evaluations on two large-scale accident datasets, namely AnAn Accident Detection (A3D) dataset and DADA-2000 dataset, validate the effectiveness of the MCD-TAD for traffic accident detection in dashcam videos.

AB - The detection of anomalies using dashcam videos is crucial for autonomous driving or driver assistance systems. The scarcity of diverse accident videos and the complex environment variations during driving significant struggle for accident detection. Leveraging the advancements in frame prediction-based accident detection methods and diffusion models, we propose a frame prediction framework based on motion-conditioned diffusion (MCD-TAD). This framework combines optical flow features with appearance features in consecutive video frames using a latent diffusion model to better capture and utilize spatiotemporal cues. Extensive evaluations on two large-scale accident datasets, namely AnAn Accident Detection (A3D) dataset and DADA-2000 dataset, validate the effectiveness of the MCD-TAD for traffic accident detection in dashcam videos.

KW - Traffic accident detection

KW - frame prediction

KW - latent diffusion model

UR - https://www.scopus.com/pages/publications/85192396871

U2 - 10.1007/978-981-97-1087-4_28

DO - 10.1007/978-981-97-1087-4_28

M3 - 会议稿件

AN - SCOPUS:85192396871

SN - 9789819710867

T3 - Lecture Notes in Electrical Engineering

SP - 292

EP - 303

BT - Proceedings of 3rd 2023 International Conference on Autonomous Unmanned Systems (3rd ICAUS 2023) - Volume III

A2 - Qu, Yi

A2 - Gu, Mancang

A2 - Niu, Yifeng

A2 - Fu, Wenxing

PB - Springer Science and Business Media Deutschland GmbH

T2 - 3rd International Conference on Autonomous Unmanned Systems, ICAUS 2023

Y2 - 9 September 2023 through 11 September 2023

ER -

Cui X, Li LL, Chai Y, Fang J, Silamu W. Self-supervised Traffic Accident Detection by Motion-Conditioned Diffusive Frame Prediction. In Qu Y, Gu M, Niu Y, Fu W, editors, Proceedings of 3rd 2023 International Conference on Autonomous Unmanned Systems (3rd ICAUS 2023) - Volume III. Springer Science and Business Media Deutschland GmbH. 2024. p. 292-303. (Lecture Notes in Electrical Engineering). doi: 10.1007/978-981-97-1087-4_28