Data Mining Practical Machine Learning Tools and Techniques

In the real world, a realistic setting for computer vision or multimedia recognition problems is that we have some classes containing lots of training data and many classes contain a small amount of training data. Therefore, how to use frequent classes to help learning rare classes for which it is harder to collect the training data is an open question. Learning with Shared Information is an emerging topic in machine learning, computer vision and multimedia analysis. There are different level of components that can be shared during concept modeling and machine learning stages, such as sharing generic object parts, sharing attributes, sharing transformations, sharing regularization parameters and sharing training examples, etc. Regarding the specific methods, multi-task learning, transfer learning and deep learning can be seen as using different strategies to share information. These learning with shared information methods are very effective in solving real-world large-scale problems. This special issue aims at gathering the recent advances in learning with shared information methods and their applications in computer vision and multimedia analysis. Both state-of-the-art works, as well as literature reviews, are welcome for submission. Papers addressing interesting real-world computer vision and multimedia applications are especially encouraged. Topics of interest include, but are not limited to:  • Multi-task learning or transfer learning for large-scale computer vision and multimedia analysis • Deep learning for large-scale computer vision and multimedia analysis • Multi-modal approach for large-scale computer vision and multimedia analysis • Different sharing strategies, e.g., sharing generic object parts, sharing attributes, sharing transformations, sharing regularization parameters and sharing training examples, • Real-world computer vision and multimedia applications based on learning with shared information, e.g., event detection, object recognition, object detection, action recognition, human head pose estimation, object tracking, location-based services, semantic indexing. • New datasets and metrics to evaluate the benefit of the proposed sharing ability for the specific computer vision or multimedia problem. • Survey papers regarding the topic of learning with shared information.  Authors who are unsure whether their planned submission is in scope may contact the guest editors prior to the submission deadline with an abstract, in order to receive feedback.

IEEE transactions on pattern analysis and machine intelligence

Modeling human behaviors and activity patterns for recognition or detection of special event has attracted significant research interest in recent years. Diverse methods that are abound for building intelligent vision systems aimed at scene understanding and making correct semantic inference from the observed dynamics of moving targets. Most applications are in surveillance, video content retrieval, and human-computer interfaces. This paper presents not only an update extending previous related surveys, but also a focus on contextual abnormal human behavior detection especially in video surveillance applications. The main purpose of this survey is to extensively identify existing methods and characterize the literature in a manner that brings key challenges to attention.

Video-Based Abnormal Human Behavior Recognition—A Review

This paper proves the uncertainty bound for the multiple Gaussian functions, termed multiple Gaussians Uncertainty (MGU), which significantly generalizes the uncertainty principle for the single Gaussian function. First, as a theoretical contribution, we prove that the momentum (velocity) and position for the sum of multiple Gaussians wave function are theoretically bounded. Second, as for a practical application, we show that the bound can be well exploited for object tracking to detect anomalies of local movement in an online learning framework. By integrating MGU with a given object tracker, we demonstrate that uncertainty principle can provide remarkable robustness in tracking. Extensive experiments are done to show that the proposed MGU can significantly help base trackers overcome the object drifting and reach state-of-the-art results.

Bounding Multiple Gaussians Uncertainty with Application to Object Tracking

This book gives a start-to-finish overview of the whole Fish4Knowledge project, in 18 short chapters, each describing one aspect of the project. The Fish4Knowledge project explored the possibilities of big video data, in this case from undersea video. Recording and analyzing 90 thousand hours of video from ten camera locations, the project gives a 3 year view of fish abundance in several tropical coral reefs off the coast of Taiwan. The research system built a remote recording network, over 100 Tb of storage, supercomputer processing, video target detection and tracking, fish species recognition and analysis, a large SQL database to record the results and an efficient retrieval mechanism. Novel user interface mechanisms were developed to provide easy access for marine ecologists, who wanted to explore the dataset. The book is a useful resource for system builders, as it gives an overview of the many new methods that were created to build the Fish4Knowledge system in a manner that also allows readers to see how all the components fit together.

/pdf/fish4knowledge-collecting-and-analyzing-massive-coral-reef-15s3qgu8ro.pdf

Fish4Knowledge: Collecting and Analyzing Massive Coral Reef Fish Video Data

We propose a novel real-time integrated unsupervised learning framework for lane detection, tracking, and road surface marking detection and recognition using live feed from a camera mounted on the dashboard of a moving vehicle. Accurate and automated lane and road surface marking detection and recognition from cameras play an important role in intelligent vehicles. It helps increase driving safety in advanced driver assistant systems and enhance global localization and path planning in autonomous driving vehicles. Map building and updation can also benefit from automated detection of lanes and road surface markings. In our framework, we propose a spatio-temporal incremental clustering algorithm coupled with curve-fitting for simultaneously detecting lanes and road surface markings. The clusters that do not represent the lanes are further processed using Grassmann manifold learning to accurately recognize the road surface markings. Our framework is independent of the number and type of lanes, be it continuous, non-continuous, straight, or curved. Moreover, the framework works for all types, sizes, and orientations of road surface markings. It is only limited by the type of road surface markings present in the training data. We have measured the performance of our framework on state-of-the-art datasets.

A Framework for Camera-Based Real-Time Lane and Road Surface Marking Detection and Recognition

In this paper, we propose a novel, real-time, camera based framework for determining the drivers emotions through facial expression recognition. Studies have established that driver's emotions play an important role in driving behavior. Therefore, continuous monitoring of the driver's emotions and requisite warning to the driver will help in maintaining safety on the roads. In our framework, at regular intervals, we detect the driver's face in the current frame and recognize the driver's emotions. For expression recognition, we extract features from the face image using two standard pre-trained deep neural networks, AlexNet and VGG16, that we fine-tune on facial expression data. We extract the features from the fully connected layer from these two networks for each frame and concatenate the two feature vectors to form a single feature vector. The novelty of our framework lies in creating distinct subspaces of each expression, using these feature vectors and applying Grassmann graph embedding based discriminant analysis to recognize the expression. The subspaces accommodate the variations in multiple instances of an expression of the same person as well as across multiple people. Our experimental results on standard datasets show that our proposed framework outperforms state-of-the-art methods.

A Framework for Driver Emotion Recognition using Deep Learning and Grassmann Manifolds

In this paper, we propose a novel, real-time driver emotion monitoring system “in the wild” based on face detection and racial expression analysis. A camera is placed inside the vehicle that continuously looks at the driver's face and monitors the driver's emotional state at regular time intervals. Camera based monitoring of the driver's attentiveness based on the driver's emotional state in naturalistic driving environments is a non-intrusive approach and an important part of an automated driver assistance system (ADAS). Our work employs a face detection model based on mixture of trees with shared pool of parts to robustly detect the drivers face in varied environmental conditions. We also extract racial landmark points, and use them to enhance our emotion recognition system. In our proposed work, we use convolution neural networks. In the first, we use VGG16 to extract appearance features from the detected face image and in the second VGG16 network, to extract geometrical features from the racial landmark points. We then combine these two features using an integration method to accurately recognize the emotions. Based on the recognized emotional state of the driver, the driver can be made aware of his emotional state in case necessary. Experimental results on publicly available driver and face expression datasets show that our system is robust and accurate for driver emotion detection.

Deep Learning Based Real-Time Driver Emotion Monitoring

In this paper, we address the problem of unsupervised learning of usual patterns of activities in an area under surveillance and detecting deviant patterns We use video epitomes for segmenting foreground objects from background and obtain an approximate shape, trajectory and temporal information in the form of space-time patches We apply pLSA for finding correlations among these patches to learn usual activities in the scene We also extend pLSA to classify a novel video as usual or unusual

/pdf/unusual-activity-analysis-using-video-epitomes-and-plsa-552pntxjs8.pdf

Unusual Activity Analysis Using Video Epitomes and pLSA

In this study, the authors propose a novel and robust approach to control auxiliary tasks in vehicles using hand gestures. First, they create a three-dimensional video volume by appending one frame to other that captures the motion history of frames. Then, they extract features using histogram of oriented gradients on each video volume. These features are represented in the form of subspaces on Grassmann manifold. To improve the recognition accuracy, they map the data from one manifold to another manifold with the help of a Grassmann kernel. Grassmann graph embedding discriminant analysis framework is used to classify the gestures. They perform experiments on two datasets: LISA and Cambridge Hand Gesture in three different testing methods such as 1/3-subject, 2/3-subject and cross-subject. Experimental results show that their proposed model outperforms and is comparable with the state-of-the-art methods.

https://ietresearch.onlinelibrary.wiley.com/doi/pdf/10.1049/iet-its.2017.0331

Ayesha Choudhary

Papers

A Framework for Camera-Based Real-Time Lane and Road Surface Marking Detection and Recognition

A Framework for Driver Emotion Recognition using Deep Learning and Grassmann Manifolds

Deep Learning Based Real-Time Driver Emotion Monitoring

Unusual Activity Analysis Using Video Epitomes and pLSA

Framework for dynamic hand gesture recognition using Grassmann manifold for intelligent vehicles