Detection of Anomalous Gait as Forensic Gait in Residential Units Using Pre-trained Convolution Neural Networks

Published on Mar 5, 2020
· DOI :10.1007/978-3-030-39442-4_57
Hana' Abd Razak (UiTM: Universiti Teknologi MARA), Ali Abd Almisreb3
Estimated H-index: 3
(International University of Sarajevo),
Nooritawati Md Tahir12
Estimated H-index: 12
(UiTM: Universiti Teknologi MARA)
One of the advantages of transfer learning technique is its capability to learn new dataset using its finest pre-trained architecture. Other advantages of this technique are small dataset requirements along with faster learning process that could yield high accuracy results. Hence in this paper, anomalous gait detection or also known as forensic gait during housebreaking crime at the gate of residential units is discussed with transfer learning technique based on five popular pre-trained convolution neural networks (CNNs) as classifiers. High accuracy and sensitivity are achieved from remodeled of the pre-trained CNNs for the learning process, offline test, and real-time test. The accuracy attained from remodeled of the pre-trained CNNs have pledged high potential towards developing the forensic intelligent surveillance technique.
  • References (49)
  • Citations (0)
📖 Papers frequently viewed together
17 Citations
2016FedCSIS: Federated Conference on Computer Science and Information Systems
1 Citations
78% of Scinapse members use related papers. After signing in, all features are FREE.
#1Nian Chi Tay (MMU: Multimedia University)H-Index: 1
#2Tee Connie (MMU: Multimedia University)H-Index: 15
Last. Pin Shen Teh (University of Manchester)H-Index: 8
view all 5 authors...
A behavior is considered abnormal when it is seen as unusual under certain contexts. The definition for abnormal behavior varies depending on situations. For example, people running in a field is considered normal but is deemed abnormal if it takes place in a mall. Similarly, loitering in the alleys, fighting or pushing each other in public areas are considered abnormal under specific circumstances. Abnormal behavior detection is crucial due to the increasing crime rate in the society. If an abn...
1 CitationsSource
#1Chaoyun ZhangH-Index: 7
#2Rui LiH-Index: 3
Last. Paul PatrasH-Index: 12
view all 5 authors...
Recognizing driver behaviors is becoming vital for in-vehicle systems that seek to reduce the incidence of car accidents rooted in cognitive distraction. In this paper, we harness the exceptional feature extraction abilities of deep learning and propose a dedicated Interwoven Deep Convolutional Neural Network (InterCNN) architecture to tackle the accurate classification of driver behaviors in real-time. The proposed solution exploits information from multi-stream inputs, i.e., in-vehicle cameras...
#1Ali Balador (MDH: Mälardalen University College)H-Index: 6
#2Anis KoubaH-Index: 1
Last. Timo Sukuvaara (Finnish Meteorological Institute)H-Index: 7
view all 13 authors...
Cooperative Cyber-Physical Systems (Co-CPSs) can be enabled using wireless communication technologies, which in principle should address reliability and safety challenges. Safety for Co-CPS enabled ...
86 CitationsSource
Recently, several techniques have been explored to detect unusual behaviour in surveillance videos. Nevertheless, few studies leverage features from pre-trained CNNs and none of then present a comparison of features generate by different models. Motivated by this gap, we compare features extracted by four state-of-the-art image classification networks as a way of describing patches from security video frames. We carry out experiments on the Ped1 and Ped2 datasets and analyze the usage of differe...
#1Jia Lu (AUT: Auckland University of Technology)
#2Wei Qi Yan (AUT: Auckland University of Technology)H-Index: 16
Last. Minh Nguyen (AUT: Auckland University of Technology)H-Index: 6
view all 3 authors...
Traditional human behaviour recognition is mostly based on global features of digital images. Nowadays, with the increase of computing power and processing capacity, deep neural networks (DNNs) acquire a high possibility to detect any objects, which have effectively led to a new era of machine learning. In this paper, we investigated a human behaviour recognition using deep learning based on YOLOv3 model. After a number of experiments conducted, our YOLOv3 model had shown to achieve 80.20% of ac...
#1Yu-Ting Pang (YZU: Yuan Ze University)
#2Sin-Wun Syu (YZU: Yuan Ze University)
Last. Bo-Hao Chen (YZU: Yuan Ze University)
view all 4 authors...
This paper introduces a trainable deep framework called DedistractedNet for recognizing the distracted driving behaviors from an image. In contrast to other conventional strategies that use physiological sensors or on-board diagnostics, the DedistractedNet directly profiles the features of driving behaviors in the image based on the deep convolutional neural networks. Experiment results manifest that the DedistractedNet achieves superior accuracy than those of other baseline CNN methods.
#1He Xu (NUPT: Nanjing University of Posts and Telecommunications)H-Index: 6
#2Leixian Shen (NUPT: Nanjing University of Posts and Telecommunications)H-Index: 1
Last. Guoxu Cao (NUPT: Nanjing University of Posts and Telecommunications)
view all 4 authors...
#1Weidong MinH-Index: 8
#2Hao CuiH-Index: 1
Last. Fangyuan ZouH-Index: 1
view all 4 authors...
1 CitationsSource
Jun 1, 2018 in CVPR (Computer Vision and Pattern Recognition)
#1Waqas Sultani (UCF: University of Central Florida)H-Index: 6
#2Chen Chen (UCF: University of Central Florida)H-Index: 55
Last. Mubarak Shah (UCF: University of Central Florida)H-Index: 88
view all 3 authors...
Surveillance videos are able to capture a variety of realistic anomalies. In this paper, we propose to learn anomalies by exploiting both normal and anomalous videos. To avoid annotating the anomalous segments or clips in training videos, which is very time consuming, we propose to learn anomaly through the deep multiple instance ranking framework by leveraging weakly labeled training videos, i.e. the training labels (anomalous or normal) are at video-level instead of clip-level. In our approach...
29 CitationsSource
In this paper, the main technologies of foreground detection, feature description and extraction, movement behavior classification and recognition were introduced. Based on optical flow for movement objects detection, optical flow energy image was put forward for movement feature expression and region convolutional neural networks was adopt to choose features and decrease dimension. Then support vector machine classifier was trained and used to classify and recognize actions. After training and ...
Cited By0