Multi-cue learning and visualization of unusual events

Schuster, René; Schulter, Samuel; Poier, Georg; Hirzer, Martin; Birchbauer, Josef Alois; Roth, Peter M.; Bischof, Horst; Winter, Martin; Schallauer, Peter

doi:10.1109/iccvw.2011.6130485

Cited by 6 publications

(3 citation statements)

References 12 publications

(16 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…the training examples regarded out-of-class, (2) and it sets a lower bound on the number of training examples used as support vectors. Again by using Lagrange techniques and using a kernel function for the dot-product calculations, the decision function becomes as in equation ( 20) and (21).…”

Section: One Class Support Vector Machinementioning

confidence: 99%

“…It loses part of the interpretability of the results as there is no restriction on the number of points that can appear on the other side of the decision boundary. Theoretically, all the points can be labelled as outlying using equation ( 20) and consequentially, the majority could have a score greater than 1.0 using equation (21). The slack variable 𝐷 𝑖 ̂ is computed using equation (24).…”

Section: One Class Support Vector Machinementioning

confidence: 99%

See 1 more Smart Citation

Hybrid Deep Learning Technique with One Class Svm for Anomaly Detection in Crowded Environment

2022

View full text Add to dashboard Cite

Nowadays, the analysis of abnormal events becomes more and more exhausting due to the divine use of surveillance cameras. Reliability of normal and abnormal samples is generally different in practice. In the existing work, developed machine learning and swarm intelligence based approaches for anomaly detection by extracting spatiotemporal features from video sequences. This proposed abnormal event detection mechanism is obvious, but the tracking result is less precise. Some of the reasons are low quality video, system noise, small object, and other factors. In order to improve the precisions of the tracked object, this research work proposed a new hybrid deep learning and robust segmentation method for better and faster tracking result. In the proposed approach, a video frame is subjected to a series of operations to extract most salient information from it. Initially, a 2D variance plane is constructed to encode local spatio-temporal variations around each pixel in a video frame. The Improved Particle Swarm Optimization algorithm is then applied to isolate the most salient regions based on motion information in the 2D variance plane. Grey Level Co-occurrence Matrix is applied to the extracted salient pixels in the video. And then the segmentation process is done with the help of fuzzy c-means clustering of successive frames are exploited for pattern matching in a simple feature space. Finally developed a hybrid deep learning based on a pre-trained Convolution Neural Network and One-class SVM is trained with spatial features for robust classification of abnormal shapes. Thus the experimental result suggests that the proposed anomaly detection techniques outperform the existing techniques in context of accuracy and time complexity.

show abstract

Section: One Class Support Vector Machinementioning

confidence: 99%

Section: One Class Support Vector Machinementioning

confidence: 99%

Hybrid Deep Learning Technique with One Class Svm for Anomaly Detection in Crowded Environment

2022

View full text Add to dashboard Cite

show abstract

“…Therefore, object recognition and tracking are beyond the scope of this paper and the details on these topics can be found in [8]. Majority of the work on abnormality detection relies on the extraction of semi-local features from video [9], [10], [11], [12], [13], that are then used to train a normalcy model. Abnormalities are detected if the normalcy model does not fit the data.…”

Section: Related Workmentioning

confidence: 99%

Video parsing for abnormality detection

Antic¹,

Ommer²

2011

2011 International Conference on Computer Vision

133

View full text Add to dashboard Cite

Detecting abnormalities in video is a challenging problem since the class of all irregular objects and behaviors isinfinite and thus no (or by far not enough) abnormal training samples are available. Consequently, a standard setting is to find abnormalities without actually knowing what they are because we have not been shown abnormal examples during training. However, although the training data does not define what an abnormality looks like, the main paradigm in this field is to directly search for individual abnormal local patches or image regions independent of another.To address this problem we parse video frames by establishing a set of hypotheses that jointly explain all the foreground while, at same time, trying to find normal training samples that explain the hypotheses. Consequently, we can avoid a direct detection of abnormalities. They are discovered indirectly as those hypotheses which are needed for covering the foreground without finding an explanation by normal samples for themselves. We present a probabilistic model that localizes abnormalities using statistical inference. On the challenging dataset of [15] it outperforms the state-of-the-art by 7% to achieve a frame-based abnormality classification performance of 91 % and the localization performance improves by 32% to 76 %.

show abstract