Enhancing human action recognition with region proposals

Deep convolutional network models have dominated recent work in human action recognition as well as image classification. However, these methods are often unduly influenced by the image background, learning and exploiting the presence of cues in typical computer vision datasets. For unbiased robotics applications, the degree of variation and novelty in action backgrounds is far greater than in computer vision datasets. To address this challenge, we propose an “action region proposal” method that, informed by optical flow, extracts image regions likely to contain actions for input into the network both during training and testing. In a range of experiments, we demonstrate that manually segmenting the background is not enough; but through active action region proposals during training and testing, state-of-the-art or better performance can be achieved on individual spatial and temporal video components. Finally, we show by focusing attention through action region proposals, we can further improve upon the existing state-of-the-art in spatio-temporally fused action recognition performance.

Human action recognition from video sequences

This PhD research has proposed new machine learning techniques to improve human action recognition based on local features. Several novel video representation and classification techniques have been proposed to increase the performance with lower computational complexity. The major contributions are the construction of new feature representation techniques, based on advanced machine learning techniques such as multiple instance dictionary learning, Latent Dirichlet Allocation (LDA) and Sparse coding. A Binary-tree based classification technique was also proposed to deal with large amounts of action categories. These techniques are not only improving the classification accuracy with constrained computational resources but are also robust to challenging environmental conditions. These developed techniques can be easily extended to a wide range of video applications to provide near real-time performance.

Multi-action recognition via stochastic modelling of optical flow and gradients

In this paper we propose a novel approach to multi-action recognition that performs joint segmentation and classification. This approach models each action using a Gaussian mixture using robust low-dimensional action features. Segmentation is achieved by performing classification on overlapping temporal windows, which are then merged to produce the final result. This approach is considerably less complicated than previous methods which use dynamic programming or computationally expensive hidden Markov models (HMMs). Initial experiments on a stitched version of the KTH dataset show that the proposed approach achieves an accuracy of 78.3%, outperforming a recent HMM-based approach which obtained 71.2%.

Kernel analysis on Grassmann manifolds for action recognition

Modelling video sequences by subspaces has recently shown promise for recognising human actions. Subspaces are able to accommodate the effects of various image variations and can capture the dynamic properties of actions. Subspaces form a non-Euclidean and curved Riemannian manifold known as a Grassmann manifold. Inference on manifold spaces usually is achieved by embedding the manifolds in higher dimensional Euclidean spaces. In this paper, we instead propose to embed the Grassmann manifolds into reproducing kernel Hilbert spaces and then tackle the problem of discriminant analysis on such manifolds. To achieve efficient machinery, we propose graph-based local discriminant analysis that utilises within-class and between-class similarity graphs to characterise intra-class compactness and inter-class separability, respectively. Experiments on KTH, UCF Sports, and Ballet datasets show that the proposed approach obtains marked improvements in discrimination accuracy in comparison to several state-of-the-art methods, such as the kernel version of affine hull image-set distance, tensor canonical correlation analysis, spatial-temporal words and hierarchy of discriminative space-time neighbourhood features.

Spatio temporal feature evaluation for action recognition

Spatio-Temporal interest points are the most popular feature representation in the field of action recognition. A variety of methods have been proposed to detect and describe local patches in video with several techniques reporting state of the art performance for action recognition. However, the reported results are obtained under different experimental settings with different datasets, making it difficult to compare the various approaches. As a result of this, we seek to comprehensively evaluate state of the art spatio- temporal features under a common evaluation framework with popular benchmark datasets (KTH, Weizmann) and more challenging datasets such as Hollywood2. The purpose of this work is to provide guidance for researchers, when selecting features for different applications with different environmental conditions. In this work we evaluate four popular descriptors (HOG, HOF, HOG/HOF, HOG3D) using a popular bag of visual features representation, and Support Vector Machines (SVM)for classification. Moreover, we provide an in-depth analysis of local feature descriptors and optimize the codebook sizes for different datasets with different descriptors. In this paper, we demonstrate that motion based features offer better performance than those that rely solely on spatial information, while features that combine both types of data are more consistent across a variety of conditions, but typically require a larger codebook for optimal performance.

Multiview gait biometrics for human identity recognition

We propose a novel multiview fusion scheme for recognizing human identity based on gait biometric data. The gait biometric data is acquired from video surveillance datasets from multiple cameras. Experiments on publicly available CASIA dataset show the potential of proposed scheme based on fusion towards development and implementation of automatic identity recognition systems.

Elliptical summary randomisation for sensor-based human activity recognition

Many conventional statistical machine learning al- gorithms generalise poorly if distribution bias ex- ists in the datasets. For example, distribution bias arises in the context of domain generalisation, where knowledge acquired from multiple source domains need to be used in a previously unseen target domains. We propose Elliptical Summary Randomisation (ESRand), an efficient domain generalisation approach that comprises of a randomised kernel and elliptical data summarisation. ESRand learns a domain interdependent projection to a la- tent subspace that minimises the existing biases to the data while maintaining the functional relationship between domains. In the latent subspace, ellipsoidal summaries replace the samples to enhance the generalisation by further removing bias and noise in the data. Moreover, the summarisation enables large-scale data processing by significantly reducing the size of the data. Through comprehensive analysis, we show that our subspace-based approach outperforms state-of-the-art results on several activity recognition benchmark datasets, while keeping the computational complexity significantly low.

Supervised latent dirichlet allocation models for efficient activity representation

Local spatio-temporal features with a Bag-of-visual words model is a popular approach used in human action recognition. Bag-of-features methods suffer from several challenges such as extracting appropriate appearance and motion features from videos, converting extracted features appropriate for classification and designing a suitable classification framework. In this paper we address the problem of efficiently representing the extracted features for classification to improve the overall performance. We introduce two generative supervised topic models, maximum entropy discrimination LDA (MedLDA) and class- specific simplex LDA (css-LDA), to encode the raw features suitable for discriminative SVM based classification. Unsupervised LDA models disconnect topic discovery from the classification task, hence yield poor results compared to the baseline Bag-of-words framework. On the other hand supervised LDA techniques learn the topic structure by considering the class labels and improve the recognition accuracy significantly. MedLDA maximizes likelihood and within class margins using max-margin techniques and yields a sparse highly discriminative topic structure; while in css-LDA separate class specific topics are learned instead of common set of topics across the entire dataset. In our representation first topics are learned and then each video is represented as a topic proportion vector, i.e. it can be comparable to a histogram of topics. Finally SVM classification is done on the learned topic proportion vector. We demonstrate the efficiency of the above two representation techniques through the experiments carried out in two popular datasets. Experimental results demonstrate significantly improved performance compared to the baseline Bag-of-features framework which uses kmeans to construct histogram of words from the feature vectors.

Adaptive unsupervised learning of human actions

Automatic detection of suspicious activities in CCTV camera feeds is crucial to the success of video surveillance systems. Such a capability can help transform the dumb CCTV cameras into smart surveillance tools for fighting crime and terror. Learning and classification of basic human actions is a precursor to detecting suspicious activities. Most of the current approaches rely on a non-realistic assumption that a complete dataset of normal human actions is available. This paper presents a different approach to deal with the problem of understanding human actions in video when no prior information is available. This is achieved by working with an incomplete dataset of basic actions which are continuously updated. Initially, all video segments are represented by Bags-Of-Words (BOW) method using only Term Frequency-Inverse Document Frequency (TF-IDF) features. Then, a data-stream clustering algorithm is applied for updating the system's knowledge from the incoming video feeds. Finally, all the actions are classified into different sets. Experiments and comparisons are conducted on the well known Weizmann and KTH datasets to show the efficacy of the proposed approach.

Understanding the very nature of Project Management : a praxiological approach

Foreword: In this paper I call upon a praxiological approach. Praxeology (early alteration of praxiology) is the study of human action and conduct. The name praxeology/praxiologyakes is root in praxis, Medieval Latin, from Greek, doing, action, from prassein to do, practice (Merriam-Webster Dictionary). Having been involved in project management education, research and practice for the last twenty years, I have constantly tried to improve and to provide a better understanding/knowledge of the field and related practice, and as a consequence widen and deepen the competencies of the people I was working with (and my own competencies as well!), assuming that better project management lead to more efficient and effective use of resources, development of people and at the end to a better world. For some time I have perceived a need to clarify the foundations of the discipline of project management, or at least elucidate what these foundations could be. An immodest task, one might say! But not a neutral one! I am constantly surprised by the way the world (i.e., organizations, universities, students and professional bodies) sees project management: as a set of methods, techniques, tools, interacting with others fields – general management, engineering, construction, information systems, etc. – bringing some effective ways of dealing with various sets of problems – from launching a new satellite to product development through to organizational change.

Composing maternal identities : the living realities of mothers with young adult-children in twenty-first century Australia

There is an abundance of books available on the topic of motherhood and mothering; the majority of these books focus on the vulnerability of babies and young children and the motherwork such vulnerability demands. In particular they focus on what it is right to do in the interests of the child, and particularly his or her growth and development. Such a focus is consistent in Western culture with modern moral frameworks where understandings of goodness have been assimilated to dimensions of human action rather than dimensions of human being, selfhood, or specific forms of life. As Charles Taylor has observed, much modern moral philosophy has focused =on what it is right to do rather than the nature of the good life‘ (1989, 13). The master narratives of motherhood and the prevailing social discourses of intensive1 and sacrificial2 mothering exemplify this view as such narratives and discourses depict =what mothers are expected to do [and] how mothers are supposed to be‘ (Nelson 2001, 140). From such infant/child-focused accounts a canonical maternal identity can be discerned; arguably, it is a restricted one. The majority of these books fail to address questions related to what it means be a mother in particular situated, existing, living realities. For instance, ask a mother with young children what being a mother means to her and she may speak of the challenges she faces balancing paid employment and her role as a mother, or the impact of the demands being made on her time and energy. However, ask a mother with young adult-children3 what being a mother means to her and she may speak in similar tones, but she may also speak in differing tones. For example, a "mature" mother may speak of the "empty nest", the "crowded house" and/or "its revolving front door". She may speak of issues related to the vulnerability of the long term marriage, elder care, or grandparenting, or even disillusionment and disenchantment. The purpose of this research is to explore the identity challenges and prospects of some mothers with young adult-children aged between 18 and 30 years of age in twenty-first century Australia. In interpreting the identity challenges and prospects this particular cohort of mothers encounter in their ordinary, everyday living, a diverse and particular range of maternal experiences.my own included5.have been traced, along with the social and ethical meanings ascribed in them. With an understanding and appreciation of voice as the medium which connects one's inner and outer worlds, this research illuminates the plurality of voices and the multiple layers of meaning in each of these mother's particular living and existing realities. Specifically, this research addresses the narrowly constructed, canonical maternal identity through a critical exploration and reflection on stories, shared in a research context, of the living realities of a group of self-identified "mature", middle-class, Australian mothers with children aged between 18 and 30 years of age6. By appraising the broader familial, historical, social, cultural, institutional, and, importantly, moral contexts in which these mothers are situated, 'thick descriptions' (Geertz 1973, 27)7 of maternal identities, and the challenges and prospects these mothers are negotiating, are provided. In terms of its ethical orientation, the frameworks which support and frame this research reject, repudiate and contest (Nelson 2001) the reduction of ethical concerns to individual or intellectual problems or dilemmas to be solved through the application of a theory derived from reasoned thinking. In dismissing deductive and =theoretical-juridical‘8 approaches, the individualistic orientation entrenched in contemporary Western moral thinking, expressed in the notion of '"what ought I to do" when faced with a problem, issue or dilemma of practical urgency' (Isaacs & Massey 1994, 1), is simultaneously rejected, repudiated and contested (Nelson 2001). In countering such understandings, this research reorients us to the illumination and articulation of who it is good to be, for each of these mothers, in allegiance with those goods which guide and inspire her orientations towards living a good life—a life which embraces and enhances the flourishing of herself and her significant others. With an understanding and appreciation that 'mind is never free of precommitment[—t]here is no innocent eye, nor is there one that penetrates aboriginal reality' (Bruner 1987, 32), this thesis is written with the voices of other interlocutors9. These interlocutors include the voices of my research participants whom I refer to as "research interlocutors", my textual "friends" — those scholars whose work resonates strongly with my orientations—as well as the myriad other voices that speak to mothers, for mothers and about mothers, such as those found in popular and mainstream press and culture. Sometimes these voices resonate; other times dissonance may be heard. In situating this research within these complementary frameworks, this research invites readers to join with me in considering, appreciating and appraising the narrow construction of maternal identity. I seek for this engagement, like the engagements with my research interlocutors, to be 'a meeting of voices, an authentic dialogue that is inclusive of the voices of all concerned participants' (Isaacs 2001, 6). I hope that the voices in this thesis resonate with yours (although, at times, you may feel some dissonance) and that together we can draw closer to the accounting, re-counting and re-stor(y)ing of maternal identities; like concentric circles of witness, the dialogue, ...will thus be expanded rippling into corners where one might both imagine, and least expect. Possibilities, then, are vast; the future exciting (Smith 2007, 397). This research is also shaped and guided by maternal scholarship, a relatively new field of inquiry known as 'motherhood studies' (O'Reilly 2011, xvii) which has its origins within the broader terrain of feminist scholarship. As a work of maternal scholarship, this thesis draws upon and continues the tradition of examining motherhood as it is experienced 'in a social context, as embedded in a political institution: in feminist terms' (Rich 1995, ix). It values mothers, their experiences, their stories, their lives. As such, this research is oriented towards 'matricentric feminism', a particular form of feminist inquiry, politics and theory which is consistent with and receptive to feminist frameworks of care and equal rights (O‘Reilly 2011, 25). A number of complementary conceptual frameworks have been engaged in this research with the thesis presented in three parts: the pre-figurative, configurative and re-configurative. As my particular living experiences provided the initial motivation for this research, an account of the challenges I experienced as a mother with young adult-children are outlined as a Prelude to this thesis. Attention then turns to Part One – Pre-figuring Maternal Identities in which the contextual, conceptual and methodological foundations underpinning this research are explored and outlined. In Chapter One, the prevailing cultural narratives and social discourses supporting and shaping the construction of the canonical maternal identity are outlined. Next, in setting the scholarly context, the critiques — arising from feminist and maternal scholarship — of motherhood as a patriarchal institution, mothering as experience, and mothering as work, are explored. As this research engaged with participants who are embedded in particular middle-class, heterosexual, familial and cultural structures, an exploration of family life cycle theory and main stream media accounts are also incorporated. The terrain in which "mature" mothering within an Australian context is experienced is also outlined, including the notions of "empty nests" and "crowded houses", grandparenting, elder care and women's midlife transition. Chapter Two gives an account of the conceptual ontological, ethical, identity and narrative frameworks underpinning this research. In setting the context for rich interpretations, the characteristics of being human10 are outlined before attention turns to our embodiment and embeddedness in our shared human condition11. From this point, attention then turns to understanding the moral form of human living12. In appreciating the vulnerability inherent in our shared human condition, the ways in which we may experience trouble in our lives is noted. The framing of identity constitution13 as complex, multi-faceted, relationally negotiated and composed is then outlined, followed by an understanding of why narrative is a valuable interpretive tool for interpreting and understanding human experiences. This chapter concludes with an appreciation of the ethical significance of storytelling. The research methodology is then outlined in Chapter Three. The rationale underpinning the adoption of the narrative interviewing technique of in-depth interviewing is explored. In exploring these methodological frameworks, the recruitment and interview processes involved in gathering and interpreting the recorded transcripts of ten Australian mothers with young adult-children are outlined. The method of analysis known as the Listening Guide14 best complements the multi-layered, pluri-vocal nature of narrative accounting. The final section of Chapter Three outlines The Guide, with one mother's recorded transcript used to illustrate this method's step-by-step process. Having gathered an understanding and appreciation of the pluri-vocal, multi-layered nature of narrative and identity constitution, the tone of this thesis changes in Part Two . Configuring Maternal Identities. This section consists of Chapters Four and Five and seeks to find meaning in, and make sense of, the differences and commonalities across these particular accounts. Chapter Four explores the living realities of four Australian mothers with young adult-children: Poppy, Honey, Lily and Heather. In presenting a thick description of these mothers' situated realities, the frameworks.the familial, social, cultural, historical and institutional backgrounds.which have supported and shaped each mother's experiences are illuminated. Simultaneously revealed through these particular accounts are the plurality of goods focusing and moving each mother to the moral form of life, a life of meaning and purpose. The harms challenging some mothers' moral motivations are also revealed in this chapter. Specifically illustrated in Chapter Four are the unique and qualitative differences of particular maternal identity configurations. Chapter Five reveals the commonalities amongst all of the research interlocutors' accounts. This chapter contests the individualistic orientation of many contemporary accounts of motherhood which are aimed at defining or contesting what a "good" mother ought to do. By turning away from such individualistic orientations, the chapter does not seek to define 'the content of obligation' (Taylor 1989, 3) but rather seeks to illuminate and articulate a richer, deeper understanding and appreciation of maternal be-ing and be-coming - that is, who it is good to be, for each of these mothers - in allegiance with those goods that focus and inspire her moral motivations. Part Three - Re-Configuring Maternal Identities, which is comprised of Chapter Six, draws this thesis to a close. In this final chapter, the preconceptions, conditions and aspirations for this mother-centred account of the living realities of a small, local cohort of mothers are reiterated. The insights gathered from the rich, descriptive accounts are illuminated and articulated, and the chapter closes with some suggestions for future research. In a Postlude, I reflect on how this research has been a transformative learning experience in my own life.an experience in which I have been able to not only deeply understand and appreciate the challenges and disorientation I was experiencing but also to identify and reorient my stance in relation to the good. In a practical sense, by offering thick descriptions of the living realities of this cohort of "mature" mothers, this research challenges the canonical maternal identity and questions its relevance for, and effect on, "mature" mothers' identity constitution. By bringing to light the complex existing realities of these particular mothers, this research critiques the canonical maternal identity by illustrating that each mother's life and her identity constitutions are complex, relationally negotiated and composed and that motherhood is an enduring way of being. Through these illustrations, this research engages with and extends understandings of difference feminism. This research, however, not only rejects, repudiates and contests (Nelson 2001) the narrowly defined canonical maternal identity. By illuminating and articulating the goods which shape and inspire these "mature" mothers' motherwork, this research offers a matricentric account which is consistent with and respectful of the particular, situated realities—the broader familial, social, institutional, but most importantly, moral values and frameworks—in which each mother‘s life is embedded and her motherwork oriented. By understanding and appreciating the complex and multiple webs of relationships in which each mother exists, this matricentric re-stor(y)ing of maternal experiences not only understands and appreciates the unique nature of each mother‘s existing realities, it is oriented to the continuing enhancing of the shared pursuit of the good which underpins particular maternal practices and particular maternal ways of being.

Governing self : SNSs as tools for self-formation

This thesis investigates how modern individuals relate to themselves and others in the service of shaping their ethical conduct and governing themselves. It considers the use of online social networking sites (SNSs) as one particular practice through which people manage their day-to-day conduct and understandings of self. Current research on the use of SNSs has conceptualised them as tools for communication, information-sharing and self-presentation. This thesis suggests a different way of thinking about these sites as tools for self-formation. A Foucaultian genealogical, historical and problematising approach is applied in order to explore processes of subjectivation and historical backgrounds involved in the use of SNSs. This is complimented with an ANT-based understanding of the role that technologies play in shaping human action. Drawing new connections between three factors will show how they contribute to the ways in which people become selves today. These factors are, one, the psychologisation and rationalisation of modern life that lead people to confess and talk about themselves in order to improve and perfect themselves, two, the transparency or publicness of modern life that incites people to reveal themselves constantly to a public audience and, three, the techno-social hybrid character of Western societies. This thesis will show how some older practices of self-formation have been translated into the context of modern technologised societies and how the care of self has been reinvigorated and combined with the notion of baring self in public. This thesis contributes a different way of thinking about self and the internet that does not seek to define what the modern self is and how it is staged online but rather accounts for the multiple, contingent and historically conditioned processes of subjectivation through which individuals relate to themselves and others in the service of governing their daily conduct.

Dictionary learning and sparse coding on Grassmann manifolds : an extrinsic solution

Recent advances in computer vision and machine learning suggest that a wide range of problems can be addressed more appropriately by considering non-Euclidean geometry. In this paper we explore sparse dictionary learning over the space of linear subspaces, which form Riemannian structures known as Grassmann manifolds. To this end, we propose to embed Grassmann manifolds into the space of symmetric matrices by an isometric mapping, which enables us to devise a closed-form solution for updating a Grassmann dictionary, atom by atom. Furthermore, to handle non-linearity in data, we propose a kernelised version of the dictionary learning algorithm. Experiments on several classification tasks (face recognition, action recognition, dynamic texture classification) show that the proposed approach achieves considerable improvements in discrimination accuracy, in comparison to state-of-the-art methods such as kernelised Affine Hull Method and graph-embedding Grassmann discriminant analysis.

Editorial of "International Journal of Critical Indigenous Studies, Volume 8, Number 2, 2015"

The articles in this edition address two critical concerns that can be broadly characterised as Indigeneity as a spectacle and the elision of Indigenous sovereignty by multiculturalism and diversity. The first article, by Maryrose Casey, examines nineteenth and early twentieth century Indigenous performances that drew on cultural practices for entertainment. She highlights how these commercially driven performances were, in fact, demonstrations of sovereignty that white colonisers paid to observe. A measure of the success of these demonstrations can be found in the reactions of audiences, which often involved disrupting the spectacle by physically occupying the performance space.