Occluded Video Instance Segmentation: A Benchmark

International Journal of Computer Vision - Can our video understanding systems perceive objects when a heavy occlusion exists in a scene? To answer this question, we collect a large-scale dataset...     

Unsupervised Learning of Probabilistic Object Models (POMs) for Object Classification, Segmentation, and Recognition Using Knowledge Propagation

We present a method to learn probabilistic object models (POMs) with minimal supervision, which exploit different visual cues and perform tasks such as classification, segmentation, and recognition. We formulate this as a structure induction and learning task and our strategy is to learn and combine elementary POMs that make use of complementary image cues. We describe a novel structure induction procedure, which uses knowledge propagation to enable POMs to provide information to other POMs and “teach them” (which greatly reduces the amount of supervision required for training and speeds up the inference). In particular, we learn a POM-IP defined on Interest Points using weak supervision [1], [2] and use this to train a POM-mask, defined on regional features, which yields a combined POM that performs segmentation/localization. This combined model can be used to train POM-edgelets, defined on edgelets, which gives a full POM with improved performance on classification. We give detailed experimental analysis on large data sets for classification and segmentation with comparison to other methods. Inference takes five seconds while learning takes approximately four hours. In addition, we show that the full POM is invariant to scale and rotation of the object (for learning and inference) and can learn hybrid objects classes (i.e., when there are several objects and the identity of the object in each image is unknown). Finally, we show that POMs can be used to match between different objects of the same category, and hence, enable objects recognition.     

An extremum principle for shape from contour

An extremum principle is developed that determines three-dimensional surface orientation from a two-dimensional contour. The principle maximizes the ratio of the area to the square of the perimeter, a measure of the compactness or symmetry of the three-dimensional surface. The principle interprets regular figures correctly and it interprets skew symmetries as oriented real symmetries. The maximum likelihood method approximates the principle on irregular figures, but we show that it consistently overestimates the slant of an ellipse.     

Recursive segmentation and recognition templates for image parsing

In this paper, we propose a Hierarchical Image Model (HIM) which parses images to perform segmentation and object recognition. The HIM represents the image recursively by segmentation and recognition templates at multiple levels of the hierarchy. This has advantages for representation, inference, and learning. First, the HIM has a coarse-to-fine representation which is capable of capturing long-range dependency and exploiting different levels of contextual information (similar to how natural language models represent sentence structure in terms of hierarchical representations such as verb and noun phrases). Second, the structure of the HIM allows us to design a rapid inference algorithm, based on dynamic programming, which yields the first polynomial time algorithm for image labeling. Third, we learn the HIM efficiently using machine learning methods from a labeled data set. We demonstrate that the HIM is comparable with the state-of-the-art methods by evaluation on the challenging public MSRC and PASCAL VOC 2007 image data sets.     

The generic viewpoint assumption and planar bias

We show that generic viewpoint and lighting assumptions resolve standard visual ambiguities by biasing toward planar surfaces. Our model uses orthographic projection with a two-dimensional affine warp and Lambertian reflectance functions, including cast and attached shadows. We use uniform priors on nuisance variables such as viewpoint direction and the light source. Limitations of using uniform priors on nuisance variables are discussed.     

Probabilistic motion estimation based on temporal coherence

We develop a theory for the temporal integration of visual motion motivated by psychophysical experiments. The theory proposes that input data are temporally grouped and used to predict and estimate the motion flows in the image sequence. This temporal grouping can be considered a generalization of the data association techniques that engineers use to study motion sequences. Our temporal grouping theory is expressed in terms of the Bayesian generalization of standard Kalman filtering. To implement the theory, we derive a parallel network that shares some properties of cortical networks. Computer simulations of this network demonstrate that our theory qualitatively accounts for psychophysical experiments on motion occlusion and motion outliers. In deriving our theory, we assumed spatial factorizability of the probability distributions and made the approximation of updating the marginal distributions of velocity at each point. This allowed us to perform local computations and simplified our implementation. We argue that these approximations are suitable for the stimuli we are considering (for which spatial coherence effects are negligible).     

Noun and adjective imagery and order in paired-associate learning by French and English subjects.

Varied rated noun imagery (I), rated adjective I, and noun-adjective order in a paired-associate (PA) learning task. 98 10th graders learned 2 PA lists, 1 with noun-adjective (N-A) pairs and 1 with adjective-noun (A-N) pairs. Within each list rated I of both nouns and adjectives was varied factorially. Both English- and French-speaking Ss learned the lists in their own languages. Results support the predictions from an hypothesis based on mediating imagery. N-A recall was superior to A-N recall for all combinations of word I, except when nouns were low and adjectives high in I. In the latter case, A-N recall was higher. This pattern was the same for both English- and French-speaking Ss. Results are interpreted as further evidence of the important role played by imagery in verbal learning. (French summary) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

The complexities of eliciting and assessing children's statements.

The authors agree with the main elements of M. Bruck and S. J. Ceci's (see record 1996-09956-001) amicus curiae brief for the case of State v. Michaels, but they believe that 4 specific aspects of child sexual abuse investigations deserve a more constructive emphasis. Research has tended to concentrate on the suggestibility of children and neglected other topics that may be of great utility to those who investigate suspected cases of child sexual abuse. More attention should have been paid in the brief to the complexities of assessment and the special characteristics of the Wee Care case. Too much emphasis has been placed on children and not nearly enough attention on remedial actions to minimize the problems associated with suggestion. Finally, research shows that it may be possible to develop reliable and valid techniques to assess child sexual abuse allegations. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Mean-field phase transitions and correlation functions for Gibbs random fields

The use of Gibbs random fields (GRF) to model images poses the important problem of the dependence of the patterns sampled from the Gibbs distribution on its parameters. Sudden changes in these patterns as the parameters are varied are known asphase transitions. In this paper we concentrate on developing a general deterministic theory for the study of phase transitions when a single parameter, namely, the temperature, is varied. This deterministic framework is based on a technique known as themean-field approximation, which is widely used in statistical physics. Our mean-field theory is general in that it is valid for any number of gray levels, any pairwise interaction potential, any neighborhood structure or size, and any set of constraints imposed on the desired images. The mean-field approximation is used to compute closed-form estimates of the critical temperatures at which phase transitions occur for two texture models widely used in the image modeling literature: the Potts model and the autobinomial model. The mean-field model allows us to gain insight into the Gibbs model behavior in the neighborhood of these temperatures. These analytical results are verified by computer simulations that use a novel mean-field descent algorithm. An important spinoff of our mean-field theory is that it allows us to compute approximations for the correlation functions of GRF models, thus bridging the gap between neighborhood-based and correlation-baseda priori image models.The work of I.M. Elfadel was supported in part by the National Science Foundation under grant MIP-91-17724. The work of A.L. Yuille was supported by the Brown, Harvard, and MIT Center for Intelligent Control Systems under U.S. Army Research Office grant DAAL03-86-C-0171, by the Defense Advanced Research Projects Agency under contract AFOSR-89-0506, and by the National Science Foundation under grant IRI-9003306.