A multi-stream CNN for deep violence detection in video sequences using handcrafted features

Mohtavipour, Seyed Mehdi; Saeidi, Mahmoud; Arabsorkhi, Abouzar

doi:10.1007/s00371-021-02266-4

Cited by 34 publications

(23 citation statements)

References 55 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, the high accuracies of 100%, 98.3%, 97.5%, 97.1%, and 96.1% were offered by the Multi-stream CNN [36], Efficient 3D CNN [37], Xception + BiLSTM + Attentions [38], AlexNet + LSTM [40] and MobileNetV2 + LSTM [34] models, respectively. In contrast, our model offers 97.62% classification accuracy for this dataset, which is 2.38% and 0.68% lower than the highest accuracies offered by Multi-stream CNN [36] and Efficient 3D CNN [37], respectively. This fairly diminished classification accuracy of our model is effectively offset by the notable improvement in total parameters reduced by 77.82% and 53.84% compared to the respective models, resulting in a favourable trade-off.…”

Section: Results and Analysismentioning

confidence: 99%

“…For higher classification accuracy, Ehasan et al [35] proposed an UNet + PatchGAN-based unsupervised action translation network utilizing spatio-temporal features to identify violent behaviours and overcome the problem related to the insufficiency of relevant data. Similarly, Mohtavipour et al [36] proposed a multi-stream CNN-based AVDC approach. Despite the promising classification performance of this model, its computational efficiency in terms of total parameters remains suboptimal.…”

Section: Spatio-temporal Feature Modelsmentioning

confidence: 99%

See 1 more Smart Citation

BrutNet: A novel approach for violence detection and classification using DCNN with GRU

Haque,

Nyeem,

Afsha

2024

The Journal of Engineering

View full text Add to dashboard Cite

Automatic Violence Detection and Classification (AVDC) with deep learning has garnered significant attention in computer vision research. This paper presents a novel approach for combining a custom Deep Convolutional Neural Network (DCNN) with a Gated Recurrent Unit (GRU) in developing a new AVDC model called BrutNet. Specifically, a time‐distributed DCNN (TD‐DCNN) is developed to generate a compact 2D representation with 512 spatial features per frame from a set of equally‐spaced frames of dimension 16090 in short video segments. Further to leverage the temporal information, a GRU layer is utilised, generating a condensed 1D vector that enables binary classification of violent or non‐violent content through multiple dense layers. Overfitting is addressed by incorporating dropout layers with a rate of 0.5, while the hidden and output layers employ rectified linear unit (ReLU) and sigmoid activations, respectively. The model is trained on the NVIDIA Tesla K80 GPU through Google Colab, demonstrating superior performance compared to existing models across various video datasets, including hockey fights, movie fights, AVD, and RWF‐2000. Notably, the model stands out by requiring only 3.416 million parameters and achieving impressive test accuracies of 97.62%, 100%, 97.22%, and 86.43% on the respective datasets. Thus, BrutNet exhibits the potential to emerge as a highly efficient and robust AVDC model in support of greater public safety, content moderation and censorship, computer‐aided investigations, and law enforcement.

show abstract

Section: Results and Analysismentioning

confidence: 99%

Section: Spatio-temporal Feature Modelsmentioning

confidence: 99%

BrutNet: A novel approach for violence detection and classification using DCNN with GRU

Haque,

Nyeem,

Afsha

2024

The Journal of Engineering

View full text Add to dashboard Cite

show abstract

“…Mohtavipour et al [22] offer a unique deep violence detection framework using characteristics extracted by manual labour. A convolutional neural network (CNN) receives these properties as streams of data in three different dimensions: space, time, and space-time.…”

Section: Related Workmentioning

confidence: 99%

“…In this study, we apply the suggested model to the UCF-Crime dataset [23], which contains a large amount of footage from public surveillance cameras documenting anomalous, unlawful, and violent behaviour in settings as diverse as schools, businesses, and streets. This dataset was chosen because its events are representative of those that occur often and in a variety of settings [24][25][26].…”

Section: Datasetmentioning

confidence: 99%

Normalized Attention Neural Network with Adaptive Feature Recalibration for Detecting the Unusual Activities Using Video Surveillance Camera

Damera¹,

Vatambeti²,

Mekala³

et al. 2023

IJSSE

View full text Add to dashboard Cite

Over the past few years, surveillance cameras have become common in many homes and businesses. Many businesses still employ a human monitor of their cameras, despite the fact that this individual is more probable to miss some anomalous occurrences in the video feeds owing to the inherent limitations of human perception. Numerous scholars have investigated surveillance data and offered several strategies for automatically identifying anomalous occurrences. Therefore, it is important to build a model for identifying unusual occurrences in the live stream from the security cameras. Recognizing potentially dangerous situations automatically so that appropriate action may be taken is crucial and can be of great assistance to law enforcement. In this research work, starting with an MRCNN for feature extraction and AFR for fine-tuning, this architecture has a number of key components (AFR). To increase the quality of the features extracted by the MRCNN, the AFR replicas the inter-dependencies among the features to enhance the quality of the low-and high-frequency features extracted. Then, a normalized attention network (NAN) is used to learn the relationships between channels, which used to identify the violence and speeds up the convergence process for training a perfect. Furthermore, the dataset took real-time security camera feeds from a variety of subjects and situations, as opposed to the hand-crafted datasets utilized in prior efforts. We also demonstrate the method's capability of assigning the correct category to each anomaly by classifying normal and abnormal occurrences. The method divided the information gathered into three primary groups: those in need of fire protection, those experiencing theft or violence, and everyone else. The study applied the proposed approach to the UCF-Crime dataset, where it outperformed other models on the same dataset.

show abstract

“…A new deep violence detection approach based on handcrafted techniques’ distinctive characteristics was presented ( Mohtavipour, Saeidi & Arabsorkhi, 2021 ). These characteristics are linked to appearance, movement speed, and representative images, and they are supplied to a CNN as spatial, temporal, and spatiotemporal streams.…”

Section: Classification Of Violence Detection Techniquesmentioning

confidence: 99%

State-of-the-art violence detection techniques in video surveillance security systems: a systematic review

Omarov

Narynov²,

Zhumanov

et al. 2022

PeerJ Computer Science

View full text Add to dashboard Cite

We investigate and analyze methods to violence detection in this study to completely disassemble the present condition and anticipate the emerging trends of violence discovery research. In this systematic review, we provide a comprehensive assessment of the video violence detection problems that have been described in state-of-the-art researches. This work aims to address the problems as state-of-the-art methods in video violence detection, datasets to develop and train real-time video violence detection frameworks, discuss and identify open issues in the given problem. In this study, we analyzed 80 research papers that have been selected from 154 research papers after identification, screening, and eligibility phases. As the research sources, we used five digital libraries and three high ranked computer vision conferences that were published between 2015 and 2021. We begin by briefly introducing core idea and problems of video-based violence detection; after that, we divided current techniques into three categories based on their methodologies: conventional methods, end-to-end deep learning-based methods, and machine learning-based methods. Finally, we present public datasets for testing video based violence detectionmethods’ performance and compare their results. In addition, we summarize the open issues in violence detection in videoand evaluate its future tendencies.

show abstract

A multi-stream CNN for deep violence detection in video sequences using handcrafted features

Cited by 34 publications

References 55 publications

BrutNet: A novel approach for violence detection and classification using DCNN with GRU

BrutNet: A novel approach for violence detection and classification using DCNN with GRU

Normalized Attention Neural Network with Adaptive Feature Recalibration for Detecting the Unusual Activities Using Video Surveillance Camera

State-of-the-art violence detection techniques in video surveillance security systems: a systematic review

Contact Info

Product

Resources

About