<p> Scene recognition performance is reduced when an observer undergoes a viewpoint shift. However, the cost of a viewpoint shift is less when it is caused by observer locomotion around a scene compared to scene rotation in front of a stationary observer- a phenomenon called the facilitative effect of locomotion. The present dissertation examined the characteristics of the facilitative effect of locomotion, and the mechanism underlying its existence. In each of six experiments, participants learned a spatial arrangement of five identical objects positioned on top of a rotatable table. Participants were then blindfolded and one object was relocated. Simultaneously, participants underwent a viewpoint shift of various magnitudes. The blindfold was then removed and participants identified which object had been moved. Chapter One showed that the facilitative effect of locomotion is robust across a wide range of viewpoint shifts (Experiment la), and that visual cues in the surrounding environment cannot account for this effect (Experiment lb). The results of Chapter Two suggest that active control over the viewpoint shift may partially account for the benefit of locomotion (Experiment 2a), specifically by providing participants with explicit knowledge regarding the magnitude and direction of the viewpoint shift (Experiment 2b ). Finally, Chapter Three showed that body-based cues available during locomotion (i .e. proprioceptive, vestibular, etc.) facilitate performance beyond actively controlling the viewpoint shift alone, and that those cues must be reliable and undisrupted to confer a scene recognition advantage (Experiment 3a). On the other hand, simply remaining oriented within one's environment could not fully account for the facilitative effect of locomotion (Experiment 3b ). These results provide an integrative account of the characteristics and mechanism associated with the facilitative effect of locomotion. Results are also discussed in the context of current views on egocentric and object-based mental transformations. </p> / Thesis / Master of Science (MSc)

Thesis advisor: Sean P. MacEvoy / For humans, healthy and productive living depends on navigating through the world and behaving appropriately along the way. But in order to do this, humans must first recognize their visual surroundings. The technical difficulty of this task is hard to comprehend: the number of possible scenes that can fall on the retina approaches infinity, and yet humans often effortlessly and rapidly recognize their surroundings. Understanding how humans accomplish this task has long been a goal of psychology and neuroscience, and more recently, has proven useful in inspiring and constraining the development of new algorithms for artificial intelligence (AI). In this thesis I begin by reviewing the current state of scene recognition research, drawing upon evidence from each of these areas, and discussing an unchallenged assumption in the literature: that scene recognition emerges from independently processing information about scenes’ local visual features (i.e. the kinds of objects they contain) and global visual features (i.e., spatial parameters. ). Over the course of several projects, I challenge this assumption with a new framework for scene recognition that indicates a crucial role for information sharing between these resources. Development and validation of this framework will expand our understanding of scene recognition in humans and provide new avenues for research by expanding these concepts to other domains spanning psychology, neuroscience, and AI. / Thesis (PhD) — Boston College, 2016. / Submitted to: Boston College. Graduate School of Arts and Sciences. / Discipline: Psychology.

Humans must regularly locate task-relevant objects when interacting with the world around them. Previous research has identified different types of information that the visual system can use to help locate objects in real-world scenes, including low-level image features and scene context. However, previous research using object arrays suggest that there may be another type of information that can guide real-world search: target knowledge. When a participant knows what a target looks like they generate and store a visual representation, or template, of it. This template then facilitates the search process. A complete understanding of real-world search needs to identify how a target template guides search through scenes. Three experiments in Chapter 2 confirmed that a target template facilitates realworld search. By using an eye-tracker target knowledge was found to facilitate both scanning and verification behaviours during search, but not the search initiation process. Within the scanning epoch a target template facilitated gaze directing and shortened fixation durations. These results suggest that target knowledge affects both the activation map, which selects which regions of the scene to fixate, and the evaluation process that compares a fixated object to the internal representation of the target. With the exact behaviours that a target template facilitates now identified, Chapter 3 investigated the role that target colour played in template-guided search. Colour is one of the more interesting target features as it has been shown to be preferred by the visual system over other features when guiding search through object arrays. Two real-world search experiments in Chapter 3 found that colour information had its strongest effect on the gaze directing process, suggesting that the visual system relies heavily on colour information when searching for target-similar regions in the scene percept. Although colour was found to facilitate the evaluation process too, both when rejecting a fixated object as a distracter and accepting it as the target, this behaviour was found to be influenced comparatively less. This suggests that the two main search behaviours – gaze directing and region evaluation – rely on different sets of template features. The gaze directing process relies heavily on colour information, but knowledge of other target features will further facilitate the evaluation process. Chapter 4 investigated how target knowledge combined with other types of information to guide search. This is particularly relevant in real-world search where several sources of guidance information are simultaneously available. A single experiment investigated how target knowledge and scene context combined to facilitate search. Both information types were found to facilitate scanning and verification behaviours. During the scanning epoch both facilitated the eye guidance and object evaluation processes. When both information sources were available to the visual system simultaneously, each search behaviour was facilitated additively. This suggests that the visual system processes target template and scene context information independently. Collectively, the results indicate not only the manner in which a target template facilitates real-world search but also updates our understanding of real-world search and the visual system. These results can help increase the accuracy of future realworld search models by specifying the manner in which our visual system utilises target template information, which target features are predominantly relied upon and how target knowledge combines with other types of guidance information.

This article describes a model for including scene/context priors in attention guidance. In the proposed scheme, visual context information can be available early in the visual processing chain, in order to modulate the saliency of image regions and to provide an efficient short cut for object detection and recognition. The scene is represented by means of a low-dimensional global description obtained from low-level features. The global scene features are then used to predict the probability of presence of the target object in the scene, and its location and scale, before exploring the image. Scene information can then be used to modulate the saliency of image regions early during the visual processing in order to provide an efficient short cut for object detection and recognition.

In the absence of cues for absolute depth measurements as binocular disparity, motion, or defocus, the absolute distance between the observer and a scene cannot be measured. The interpretation of shading, edges and junctions may provide a 3D model of the scene but it will not inform about the actual "size" of the space. One possible source of information for absolute depth estimation is the image size of known objects. However, this is computationally complex due to the difficulty of the object recognition process. Here we propose a source of information for absolute depth estimation that does not rely on specific objects: we introduce a procedure for absolute depth estimation based on the recognition of the whole scene. The shape of the space of the scene and the structures present in the scene are strongly related to the scale of observation. We demonstrate that, by recognizing the properties of the structures present in the image, we can infer the scale of the scene, and therefore its absolute mean depth. We illustrate the interest in computing the mean depth of the scene with application to scene recognition and object detection.

No description available.

This article describes a model for including scene/context priors in attention guidance. In the proposed scheme, visual context information can be available early in the visual processing chain, in order to modulate the saliency of image regions and to provide an efficient short cut for object detection and recognition. The scene is represented by means of a low-dimensional global description obtained from low-level features. The global scene features are then used to predict the probability of presence of the target object in the scene, and its location and scale, before exploring the image. Scene information can then be used to modulate the saliency of image regions early during the visual processing in order to provide an efficient short cut for object detection and recognition.

Is there a significant difference in the design of deep networks for the tasks of classifying object-centric images and scenery images? How to design networks that extract the most representative features for scene recognition? To answer these questions, we design studies to examine the scales and richness of image features for scenery image recognition. Three methods are proposed that integrate the scale factor to the deep networks and reveal the fundamental network design strategies. In our first attempt to integrate scale factors into the deep network, we proposed a method that aggregates both the context and multi-scale object information of scene images by constructing a multi-scale pyramid. In our design, integration of object-centric multi-scale networks achieved a performance boost of 9.8%; integration of object- and scene-centric models obtained an accuracy improvement of 5.9% compared with single scene-centric models. We also exploit bringing the attention scheme to the deep network and proposed a Scale Attentive Network (SANet). The SANet streamlines the multi-scale scene recognition pipeline, learns comprehensive scene features at various scales and locations, addresses the inter-dependency among scales, and further assists feature re-calibration as well as the aggregation process. The proposed network achieved a Top-1 accuracy increase by 1.83% on Place365 standard dataset with only 0.12% additional parameters and 0.24% additional GFLOPs using ResNet-50 as the backbone. We further bring the scale factor implicitly into network backbone design by proposing a Deep-Narrow Network and Dilated Pooling module. The Deep-narrow architecture increased the depth of the network as well as decreased the width of the network, which uses a variety of receptive fields by stacking more layers. We further proposed a Dilated Pooling module which expanded the pooling scope and made use of multi-scale features in the pooling operation. By embedding the Dilated Pooling into Deep-Narrow Network, we obtained a Top-1 accuracy boost of 0.40% using less than half of the GFLOPs and parameters compared to benchmark ResNet-50.

This thesis attempts to develop features identification and tracking system for an autonomous ground vehicle by focusing on four fundamental tasks: Motion detection, object tracking, scene recognition, and object detection and recognition. For motion detection, we combined the background subtraction method using the mixture of Gaussian models and the optical flow to highlight any moving objects or new entering objects which stayed still. To increase robustness for object tracking result, we used the Kalman filter to combine the tracking method based on the color histogram and the method based on invariant features. For scene recognition, we applied the algorithm Census Transform Histogram (CENTRIST), which is based on Census Transform images of the training data and the Support Vector Machine classifier, to recognize a total of 8 scene categories. Because detecting the horizon is also an important task for many navigation applications, we also performed horizon detection in this thesis. Finally, the deformable parts-based models algorithm was implemented to detect some common objects, such as humans and vehicles. Furthermore, objects were only detected in the area under the horizon to reduce the detecting time and false matching rate. / Master of Science

This thesis is concerned with the problem of situational awareness in autonomous vehicles. In this context, situational awareness refers to the ability of an autonomous vehicle to perceive the road layout ahead, interpret the implied semantics and gain an awareness of its surrounding - thus reading the road ahead. Autonomous vehicles require a high level of situational awareness in order to operate safely and efficiently in real-world dynamic environments. A system is therefore needed that is able to model the expected road layout in terms of semantics, both under normal and roadwork conditions. This thesis takes a three-pronged approach to this problem: Firstly, we consider reading the road surface. This is formulated in terms of probabilistic road marking classification and interpretation. We then derive the road boundaries using only a 2D laser and algorithms based on geometric priors from Highway Traffic Engineering principles. Secondly, we consider reading the road scene. Here, we formulate a roadwork scene recognition framework based on opponent colour vision in humans. Finally, we provide a data representation for situational awareness that unifies reading the road surface and reading the road scene. This thesis therefore frames situational awareness in autonomous vehicles in terms of both static and dynamic road semantics - and detailed formulations and algorithms are discussed. We test our algorithms on several benchmarking datasets collected using our autonomous vehicle on both rural and urban roads. The results illustrate that our road boundary estimation, road marking classification, and roadwork scene recognition frameworks allow autonomous vehicles to truly and meaningfully read the semantics of the road ahead, thus gaining a valuable sense of situational awareness even at challenging layouts, roadwork sites, and along unknown roadways.