CADP: A Novel Dataset for CCTV Traffic Camera based Accident Analysis

Shah, Ankit; Lamare, Jean-Bapstite; Nguyen-Anh, Tuan; Hauptmann, Alexander G.

doi:10.1109/avss.2018.8639160

Cited by 83 publications

(33 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In this method, Histogram of gradients (HOG) and Gray level co-occurrence matrix features were used to train support vector machines. Reference [30] presented a novel dataset for car accidents analysis based on traffic Closed-Circuit Television (CCTV) footage, and combined Faster Regions-Convolutional Neural Network (R-CNN) and Context Mining to detect and predict car accidents. The method in [30] achieved 1.68 seconds in terms of Time-To-Accident measure with an Average Precision of 47.25%.…”

Section: B Methods Based Video Featuresmentioning

confidence: 99%

“…Reference [30] presented a novel dataset for car accidents analysis based on traffic Closed-Circuit Television (CCTV) footage, and combined Faster Regions-Convolutional Neural Network (R-CNN) and Context Mining to detect and predict car accidents. The method in [30] achieved 1.68 seconds in terms of Time-To-Accident measure with an Average Precision of 47.25%. Reference [8] proposed a novel framework for automatic car accident detection, which learned feature representation from the spatio-temporal volumes of raw pixel intensity instead of traditional hand-crafted features.…”

Section: B Methods Based Video Featuresmentioning

confidence: 99%

See 1 more Smart Citation

An Automatic Car Accident Detection Method Based on Cooperative Vehicle Infrastructure Systems

et al. 2019

View full text Add to dashboard Cite

Car accidents cause a large number of deaths and disabilities every day, a certain proportion of which result from untimely treatment and secondary accidents. To some extent, automatic car accident detection can shorten response time of rescue agencies and vehicles around accidents to improve rescue efficiency and traffic safety level. In this paper, we proposed an automatic car accident detection method based on Cooperative Vehicle Infrastructure Systems (CVIS) and machine vision. First of all, a novel image dataset CAD-CVIS is established to improve accuracy of accident detection based on intelligent roadside devices in CVIS. Especially, CAD-CVIS is consisted of various kinds of accident types, weather conditions and accident location, which can improve self-adaptability of accident detection methods among different traffic situations. Secondly, we develop a deep neural network model YOLO-CA based on CAD-CVIS and deep learning algorithms to detect accident. In the model, we utilize Multi-Scale Feature Fusion (MSFF) and loss function with dynamic weights to enhance performance of detecting small objects. Finally, our experiment study evaluates performance of YOLO-CA for detecting car accidents, and the results show that our proposed method can detect car accident in 0.0461 seconds (21.6FPS) with 90.02% average precision (AP). In additionally, we compare YOLO-CA with other object detection models, and the results demonstrate the comprehensive performance improvement on the accuracy and real-time over other models.INDEX TERMS Car accident detection, CVIS, machine vision, deep learning.

show abstract

Section: B Methods Based Video Featuresmentioning

confidence: 99%

Section: B Methods Based Video Featuresmentioning

confidence: 99%

An Automatic Car Accident Detection Method Based on Cooperative Vehicle Infrastructure Systems

et al. 2019

View full text Add to dashboard Cite

show abstract

“…The Car Accident Detection and Prediction (CADP) [14] contains traffic collision videos collected from Youtube. We will use it to test the performance on traffic collision detection.…”

Section: ) Dataset and Methodsmentioning

confidence: 99%

Training-free Monocular 3D Event Detection System for Traffic Surveillance

Chen

Liu

et al. 2019

2019 IEEE International Conference on Big Data (Big Data)

View full text Add to dashboard Cite

We focus on the problem of detecting traffic events in a surveillance scenario, including the detection of both vehicle actions and traffic collisions. Existing event detection systems are mostly learning-based and have achieved convincing performance when a large amount of training data is available. However, in real-world scenarios, collecting sufficient labeled training data is expensive and sometimes impossible (e.g. for traffic collision detection). Moreover, the conventional 2D representation of surveillance views is easily affected by occlusions and different camera views in nature. To deal with the aforementioned problems, in this paper, we propose a training-free monocular 3D event detection system for traffic surveillance. Our system firstly projects the vehicles into the 3D Euclidean space and estimates their kinematic states. Then we develop multiple simple yet effective ways to identify the events based on the kinematic patterns, which need no further training. Consequently, our system is robust to the occlusions and the viewpoint changes. Exclusive experiments report the superior result of our method on large-scale real-world surveillance datasets, which validates the effectiveness of our proposed system. The demonstration videos of our system are available online 1 .

show abstract

“…They utilized features of candidate objects as spatial stream and optical flow as temporal stream, and the two-stream features are fused for risk level classification. Instead of using dashcam videos, Shah et al [30] proposed to use surveillance videos to anticipate traffic accidents by using the framework DSA. Different from previous works, recently Neumann and Zisserman [25] used 3D convolutional networks to predict the sufficient statistics of a mixture of 1D Gaussian distributions.…”

Section: Related Work 21 Traffic Accident Anticipationmentioning

confidence: 99%

“…Graph representation for traffic scene has the advantages over full-frame feature embedding in that the impact of cluttered traffic background can be reduced and informative relations of traffic agents can be discovered for accident anticipation. Similar to [4,30], we exploit object detectors [3,27] to obtain a fixed number of candidate objects. These objects are treated as graph nodes so that a complete graph can be formed.…”

Section: Spatio-temporal Relational Learningmentioning

confidence: 99%

Uncertainty-based Traffic Accident Anticipation with Spatio-Temporal Relational Learning

Bao

Kong

2020

Proceedings of the 28th ACM International Conference on Multimedia

View full text Add to dashboard Cite

Time-to-Accident threshold Figure 1: Illustration of Uncertainty-based Accident Anticipation. This paper presents a novel model to predict the probabilities (black curve) of a future accident (ranges from 90-th to 100-th frame). Our goal is to achieve early anticipation (large Time-to-Accident) giving a threshold probability (horizontal dashed line), while estimating two kinds of predictive uncertainties, i,e., aleatoric uncertainty (wheat color region) and epistemic uncertainty (blue region).

show abstract

CADP: A Novel Dataset for CCTV Traffic Camera based Accident Analysis

Cited by 83 publications

References 14 publications

An Automatic Car Accident Detection Method Based on Cooperative Vehicle Infrastructure Systems

An Automatic Car Accident Detection Method Based on Cooperative Vehicle Infrastructure Systems

Training-free Monocular 3D Event Detection System for Traffic Surveillance

Uncertainty-based Traffic Accident Anticipation with Spatio-Temporal Relational Learning

Contact Info

Product

Resources

About