The seminal multiple view stereo benchmark evaluations from Middlebury and by Strecha et al. have played a major role in propelling the development of multi-view stereopsis methodology. Although seminal, these datasets are limited scope with few reference scenes. Here, we try to take works step further proposing new dataset, which is an order magnitude larger number scenes significant increase diversity. Specifically, propose dataset containing 80 large variability. Each scene consists 49 or...

This paper addresses the problem of obtaining complete, detailed reconstructions textureless shiny objects. We present an algorithm which uses silhouettes object, as well images obtained under changing illumination conditions. In contrast with previous photometric stereo techniques, ours is not limited to a single viewpoint but produces accurate in full 3D. A number object are from multiple viewpoints, varying lighting Starting silhouettes, recovers camera motion and constructs object's...

This paper presents a volumetric formulation for the multi-view stereo problem which is amenable to computationally tractable global optimisation using Graph-cuts. Our approach seek optimal partitioning of 3D space into two regions labelled as "object" and "empty" under cost functional consisting following terms: (1) A term that forces boundary between pass through photo-consistent locations (2) ballooning inflates region. To take account effect occlusion on first we use an robust...

This paper presents a novel formulation for the multi-view scene reconstruction problem. While this benefits from volumetric representation, it is amenable to computationally tractable global optimisation using Graph-cuts. The algorithm proposed uses visual hull of infer occlusions and as constraint on topology scene. A photo consistency-based surface cost functional defined discretised with weighted graph. optimal under obtained minimum cut solution Our method provides viewpoint independent...

We present an algorithm and the associated capture methodology to acquire track detailed 3D shape, bends, wrinkles of deforming surfaces. Moving data has been difficult obtain by methods that rely on known surface features, structured light, or silhouettes. Multispec- tral photometric stereo is attractive alternative because it can recover a dense normal field from un-textured surface. show how such register over time generate single Experiments were performed video sequences un- textured...

Current, self-supervised depth estimation architectures rely on clear and sunny weather scenes to train deep neural networks. However, in many locations, this assumption is too strong. For example the UK (2021), 149 days consisted of rain. these be effective real-world applications, we must create models that can generalise all conditions, times day image qualities. Using a combination computer graphics generative models, one augment existing sunny-weather data variety ways simulate adverse...

We present a new formulation to multi-view stereo that treats the problem as probabilistic 3D segmentation. Previous work has used photo-consistency criterion detector of boundary between scene and surrounding empty space. Here we show how same can also provide foreground/background model predict if location is inside or outside scene. This replaces commonly naive foreground based on ballooning which known perform poorly in concavities. demonstrate visibility linked previous depth-map fusion...

This paper concerns the problem of agent trust in an electronic market place. We maintain that involves making decisions under uncertainty and therefore phenomenon should be modelled probabilistically. propose a probabilistic framework models interactions as Hidden Markov Model (HMM). The observations HMM are interaction outcomes hidden state is underlying probability good outcome. task deciding whether to interact with another reduces inference current given all previous outcomes. model...

In urban areas, the efficiency of traffic flows largely depends on signal operation and expansion existing infrastructure is not feasible due to spatial, economic environmental constraints. this paper, we address problem congestion around road intersections. We developed our simulator optimally simulate various scenarios, closely related real-world situations. contend that adaptive real-time optimization key improving infrastructure's effectiveness by enabling control system learn, adapt...

We present an algorithm and the associated single-view capture methodology to acquire detailed 3D shape, bends, wrinkles of deforming surfaces. Moving data has been difficult obtain by methods that rely on known surface features, structured light, or silhouettes. Multispectral photometric stereo is attractive alternative because it can recover a dense normal field from untextured surface. show how such data, which in turn allows us demonstrate strengths limitations our simple frame-to-frame...

This paper addresses the challenge of reading low contrast text on tyre sidewall images vehicles in motion. It presents first its kind, a full scale industrial system which can read codes when installed along driveways such as at gas stations or parking lots with driving under 10 mph. Tyre circularity is detected using circular Hough transform dynamic radius detection. The arches are then unwarped into rectangular patches. A cascade convolutional neural network (CNN) classifiers applied for...

This paper addresses the problem of obtaining complete, detailed reconstructions shiny textureless objects. We present an algorithm which uses silhouettes object, as well images obtained under varying illumination conditions. In contrast with previous photometric stereo techniques, ours is not limited to a single viewpoint and produces accurate in full 3D. A number object are from multiple viewpoints, lighting Starting silhouettes, recovers camera motion constructs object's visual hull. then...

This paper addresses the problem of automatically obtaining object/background segmentation a rigid 3D object observed in set images that have been calibrated for camera pose and intrinsics. Such segmentations can be used to obtain shape representation potentially texture-less by computing visual hull. We propose an automatic approach where segmented is identified cameras instead user input such as 2D bounding rectangles or brush-strokes. The key behind our method pairwise MRF framework...

Light occlusions are one of the most significant difficulties photometric stereo methods. When three or more images available without occlusion, local surface orientation is overdetermined so that shape can be computed and shadowed pixels discarded. In this paper, we look at challenging case when only two leading to a degree freedom ambiguity per pixel in orientation. We show that, presence noise, integrability alone cannot resolve reconstruct geometry regions. As problem ill-posed describe...

This work proposes a system for retrieving clothing and fashion products from video content. Although films television are the perfect showcase brands to promote their products, spectators not always aware of where buy latest trends they see on screen. Here, framework breaking gap between shown videos users is presented. By relating items frames in an indexed database performing frame retrieval with temporal aggregation fast indexing techniques, we can find simple non-intrusive way....

We propose an algorithm for automatically obtaining a segmentation of rigid object in sequence images that are calibrated camera pose and intrinsic parameters. Until recently, the best results have been obtained by interactive methods require manual labelling image regions. Our method requires no user input but instead relies on fixating interest during sequence. begin learning model is colour, from pixels around fixation points. then extract edges combine these with colour information...