Search

Visual-inertial odometry (VIO) is the pose estimation backbone for most AR/VR and autonomous robotic systems today, in both academia and industry. However, these systems are highly sensitive to the initialization of key parameters such as sensor biases, gravity direction, and metric scale. In practical scenarios where high-parallax or variable acceleration assumptions are rarely met (e.g. hovering aerial robot, smartphone AR user not gesticulating with phone), classical visual-inertial initialization formulations often become ill-conditioned and/or fail to meaningfully converge. In this paper we target visual-inertial initialization specifically for these low-excitation scenarios critical to in-the-wild usage. We propose to circumvent the limitations of classical visual-inertial structure-from-motion (SfM) initialization by incorporating a new learning-based measurement as a higher-level input. We leverage learned monocular depth images (mono-depth) to constrain the relative depth of features, and upgrade the mono-depths to metric scale by jointly optimizing for their scales and shifts. Our experiments show a significant improvement in problem conditioning compared to a classical formulation for visual-inertial initialization, and demonstrate significant accuracy and robustness improvements relative to the state-of-the-art on public benchmarks, particularly under low-excitation scenarios. We further extend this improvement to implementation within an existing odometry system to illustrate the impact of our improved initialization method on resulting tracking trajectories., Comment: to be published in 2022 European Conference on Computer Vision

In this dissertation, we present several techniques used to automatically generate virtual models of indoor building environments. The interior environment of a building is scanned by a custom hardware system, which provides raw laser and camera sensor readings used to develop these models. Our modeling techniques can be separated into three categories: 2D floor plan models, simplified 2.5D extruded models, and fully complex and detailed 3D models. All models are produced automatically from the output data of a backpack-mounted ambulatory scanning system, which can scan multiple floors of a building efficiently. We can capture the entirety of a large building in only a few hours, including staircases and other hard to reach locations, which improves upon the state-of-the-art static scanning by several orders of magnitude. Our algorithms are capable of producing many kinds of 3D virtual building models while traversing through a GPS-denied environment, including point clouds, 2D floor plans, and 3D building models.The novel contributions of this dissertation fall into three groups. First, we present multiple methods for producing 2D floor plans in a scalable fashion. We automatically partition these floor plans into separate building levels and separate rooms within each level. These floor plans can also be extruded into simplified 2.5D models of the building environment. Second, we present multiple techniques to produce complex 3D models of the scanned environment, capturing all observed detail. Third, we present several techniques that combine these two types of building models -- simple and complex -- to perform additional analysis of the building environment. Such analysis includes segmenting objects and furniture in the environment as separate models, reducing noise and artifacts within the models, and demonstrating novel visualization techniques. Additionally, we show several example datasets generated by the system in real-world environments, including buildin

In this dissertation, we present several techniques used to automatically generate virtual models of indoor building environments. The interior environment of a building is scanned by a custom hardware system, which provides raw laser and camera sensor readings used to develop these models. Our modeling techniques can be separated into three categories: 2D floor plan models, simplified 2.5D extruded models, and fully complex and detailed 3D models. All models are produced automatically from the output data of a backpack-mounted ambulatory scanning system, which can scan multiple floors of a building efficiently. We can capture the entirety of a large building in only a few hours, including staircases and other hard to reach locations, which improves upon the state-of-the-art static scanning by several orders of magnitude. Our algorithms are capable of producing many kinds of 3D virtual building models while traversing through a GPS-denied environment, including point clouds, 2D floor plans, and 3D building models.The novel contributions of this dissertation fall into three groups. First, we present multiple methods for producing 2D floor plans in a scalable fashion. We automatically partition these floor plans into separate building levels and separate rooms within each level. These floor plans can also be extruded into simplified 2.5D models of the building environment. Second, we present multiple techniques to produce complex 3D models of the scanned environment, capturing all observed detail. Third, we present several techniques that combine these two types of building models -- simple and complex -- to perform additional analysis of the building environment. Such analysis includes segmenting objects and furniture in the environment as separate models, reducing noise and artifacts within the models, and demonstrating novel visualization techniques. Additionally, we show several example datasets generated by the system in real-world environments, including buildin

We report on wide-field optically detected magnetic resonance imaging of nitrogen-vacancy centers (NVs) in type IIa polycrystalline diamond. These studies reveal a heterogeneous crystalline environment that produces a varied density of NV centers, including preferential orientation within some individual crystal grains, but preserves long spin coherence times. Using the native NVs as nanoscale sensors, we introduce a three-dimensional strain imaging technique with high sensitivity (<10[superscript -5] Hz[superscript –1/2]) and diffraction-limited resolution across a wide field of view., Allen Institute for Brain Science, Howard Hughes Medical Institute

This dissertation project focuses on dynamics of mobilization and repression in the Occupy movement. This movement emerged in late 2011 in the United States and sought to protest the growing economic inequality and the growing influence of corporations in politics. This project focuses on Occupy chapters which emerged in 74 out of the 100 largest cities in the United States. The first empirical chapter of this project focuses on dynamics which affect differences in protest size, measured in terms of protest turnouts relative to population. This chapter first demonstrates the importance of large student populations and greater numbers of universities in making large turnouts more likely, then focuses on accounting for the aspects of student populations and colleges which play a role in affecting protest size. The findings show that larger protests are more likely in cities with smaller, more liberal colleges, but also with low academic rankings, with institutional support from Chicano Studies and with proportions of economically disadvantaged students. The second part of the dissertation focuses on duration. In particular, this chapter seeks to explain the causes behind relative levels of duration of Occupy protest encampments. The findings show that protest size has an inverse relationship with protest duration: therefore, smaller movements are more likely to last. The findings also show that Occupy chapters can last longer by retaining key logistic resources as well as avoiding elected leader criticism in the media. The third and final empirical chapter of this dissertation focuses on the role of the media in making violent repression more likely. This chapter first analyzes the role of different types of threat (including protest turnouts and protester violence and property damage) in making overall, positive and negative media coverage more or less likely. The findings show that all types of threat have a positive effect on positive and negative coverage but o

Hmx1 encodes a homeodomain transcription factor expressed in the developing lateral craniofacial mesenchyme, retina and sensory ganglia. Mutation or mis-regulation of Hmx1 underlies malformations of the eye and external ear in multiple species. Deletion or insertional duplication of an evolutionarily conserved region (ECR) downstream of Hmx1 has recently been described in rat and cow, respectively. Here, we demonstrate that the impact of Hmx1 loss is greater than previously appreciated, with a variety of lateral cranioskeletal defects, auriculofacial nerve deficits, and duplication of the caudal region of the external ear. Using a transgenic approach, we demonstrate that a 594 bp sequence encompassing the ECR recapitulates specific aspects of the endogenous Hmx1 lateral facial expression pattern. Moreover, we show that Hoxa2, Meis and Pbx proteins act cooperatively on the ECR, via a core 32 bp sequence, to regulate Hmx1 expression. These studies highlight the conserved role for Hmx1 in BA2-derived tissues and provide an entry point for improved understanding of the causes of the frequent lateral facial birth defects in humans.

Hmx1 encodes a homeodomain transcription factor expressed in the developing lateral craniofacial mesenchyme, retina and sensory ganglia. Mutation or mis-regulation of Hmx1 underlies malformations of the eye and external ear in multiple species. Deletion or insertional duplication of an evolutionarily conserved region (ECR) downstream of Hmx1 has recently been described in rat and cow, respectively. Here, we demonstrate that the impact of Hmx1 loss is greater than previously appreciated, with a variety of lateral cranioskeletal defects, auriculofacial nerve deficits, and duplication of the caudal region of the external ear. Using a transgenic approach, we demonstrate that a 594 bp sequence encompassing the ECR recapitulates specific aspects of the endogenous Hmx1 lateral facial expression pattern. Moreover, we show that Hoxa2, Meis and Pbx proteins act cooperatively on the ECR, via a core 32 bp sequence, to regulate Hmx1 expression. These studies highlight the conserved role for Hmx1 in BA2-derived tissues and provide an entry point for improved understanding of the causes of the frequent lateral facial birth defects in humans.

The scope of this thesis is to analyze the feasibility of having different wireless mesh network architectures transfer data to a wired network via a joint (universal) access point (UAP). Additionally this thesis analyzes the feasibility of using similar joint (universal) access point technology to allow heterogeneous wireless mesh network devices in close proximally to the UAP transmit data to/from each other via the UAP. This research also includes evaluating COTS tools for possible implementation of a joint access point as well as seeking partnership with private industry to assist in research efforts and/or the development or joint (universal) access point solution(s). The thesis concludes with a recommendation on application of universal joint point technology, to include recommendations for implementation of such technology in the Tactical Network Topology (TNT) environment., http://archive.org/details/solutionnalysiso109452951, Approved for public release; distribution is unlimited.

Sensory and autonomic neurons of the vertebrate peripheral nervous system are derived from the neural crest. Here we use the expression of lineage-specific transcription factors as a means to identify neuronal subtypes that develop in rat neural crest cultures grown in a defined medium. Sensory neurons, identified by expression of the POU-domain transcription factor Brn-3.0, develop from dividing precursors that differentiate within 2 days following emigration from the neural tube. Most of these precursors generate sensory neurons even when challenged with BMP2, a factor that induces autonomic neurogenesis in many other cells in the explants. Moreover, BMP2 fails to prevent expression of the sensory-specific basic helix-loop-helix (bHLH) transcription factors neurogenin1, neurogenin2 and neuroD, although it induces expression of the autonomic-specific bHLH factor MASH1 and the paired homeodomain factor Phox2a in other cells. These data suggest that there are mitotically active precursors in the mammalian neural crest that can generate sensory neurons even in the presence of a strong autonomic-inducing cue. Further characterization of the neurons generated from such precursors indicates that, under these culture conditions, they exhibit a proprioceptive and/or mechanosensory, but not nociceptive, phenotype. Such precursors may therefore correspond to a lineally (Frank, E. and Sanes, J. (1991) Development 111, 895-908) and genetically (Ma, Q., Fode, C., Guillemot, F. and Anderson, D. J. (1999) Genes Dev. 13, in press) distinct subset of early-differentiating precursors of large-diameter sensory neurons identified in vivo.

Sensory and autonomic neurons of the vertebrate peripheral nervous system are derived from the neural crest. Here we use the expression of lineage-specific transcription factors as a means to identify neuronal subtypes that develop in rat neural crest cultures grown in a defined medium. Sensory neurons, identified by expression of the POU-domain transcription factor Brn-3.0, develop from dividing precursors that differentiate within 2 days following emigration from the neural tube. Most of these precursors generate sensory neurons even when challenged with BMP2, a factor that induces autonomic neurogenesis in many other cells in the explants. Moreover, BMP2 fails to prevent expression of the sensory-specific basic helix-loop-helix (bHLH) transcription factors neurogenin1, neurogenin2 and neuroD, although it induces expression of the autonomic-specific bHLH factor MASH1 and the paired homeodomain factor Phox2a in other cells. These data suggest that there are mitotically active precursors in the mammalian neural crest that can generate sensory neurons even in the presence of a strong autonomic-inducing cue. Further characterization of the neurons generated from such precursors indicates that, under these culture conditions, they exhibit a proprioceptive and/or mechanosensory, but not nociceptive, phenotype. Such precursors may therefore correspond to a lineally (Frank, E. and Sanes, J. (1991) Development 111, 895-908) and genetically (Ma, Q., Fode, C., Guillemot, F. and Anderson, D. J. (1999) Genes Dev. 13, in press) distinct subset of early-differentiating precursors of large-diameter sensory neurons identified in vivo.

This is a work-in-progress. It backgrounds the Loyal Orange institution from its inception in Ireland, its early involvement in colonial politics and its tumultuous establishment in Sydney in 1845 and the entry of women into its lodges during the 19th century.