No description available.

Human performance in object recognition changes with practice, even in the absence of feedback to the subject. The nature of the change can reveal important properties of the process of recognition. We report an experiment designed to distinguish between non-specific task learning and object- specific practice effects. The results of the experiment support the notion that learning through modification of object representations can be separated from less interesting effects of practice, if appropriate response measures (specifically, the coefficient of variation of response time over views of an object) are used. Furthermore, the results, obtained with computer-generated amoeba-like objects, corroborate previous findings regarding the development of canonical views and related phenomena with practice.

Image representation is a key component in visual recognition systems. In visual recognition problem, the solution or the model should be able to learn and infer the quality of certain visual semantics in the image. Therefore, it is important for the model to represent the input image in a way that the semantics of interest can be inferred easily and reliably. This thesis is written in the form of a compilation of publications and tries to look into the Convolutional Networks (CovnNets) representation in visual recognition problems from an empirical perspective. Convolutional Network is a special class of Neural Networks with a hierarchical structure where every layer’s output (except for the last layer) will be the input of another one. It was shown that ConvNets are powerful tools to learn a generic representation of an image. In this body of work, we first showed that this is indeed the case and ConvNet representation with a simple classifier can outperform highly-tuned pipelines based on hand-crafted features. To be precise, we first trained a ConvNet on a large dataset, then for every image in another task with a small dataset, we feedforward the image to the ConvNet and take the ConvNets activation on a certain layer as the image representation. Transferring the knowledge from the large dataset (source task) to the small dataset (target task) proved to be effective and outperformed baselines on a variety of tasks in visual recognition. We also evaluated the presence of spatial visual semantics in ConvNet representation and observed that ConvNet retains significant spatial information despite the fact that it has never been explicitly trained to preserve low-level semantics. We then tried to investigate the factors that affect the transferability of these representations. We studied various factors on a diverse set of visual recognition tasks and found a consistent correlation between the effect of those factors and the similarity of the target task to the source task. This intuition alongside the experimental results provides a guideline to improve the performance of visual recognition tasks using ConvNet features. Finally, we addressed the task of visual instance retrieval specifically as an example of how these simple intuitions can increase the performance of the target task massively. / <p>QC 20161209</p>

Existing computer programs designed to perform visual recognition of objects suffer from a basic weakness: the inability to spotlight regions in the image that potentially correspond to objects of interest. The brain's mechanisms of visual attention, elucidated by psychophysicists and neurophysiologists, may suggest a solution to the computer's problem of object recognition.

As life expectancy increases in modern societies, a greater importance has recently started to be given to cognitive aging. Alzheimer's disease (AD) affects the memory capability of individuals at advanced ages, independently of their general physical health. However, AD is suggested to have an undetectable development many years prior the first clear behavioral symptoms. This silent presence of AD may allow scientists to detect its initial stages, at which a combination of prevention treatments, such as medication and cognitive training, can be more effective. This study extends a line of research that aims to identify possible 'silent' biomarkers of AD using working memory performance and electrophysiological recordings (EEG) in healthy adults. Working memory (aka., short-term memory) is a memory sub-type used in everyday life that allows us to execute tasks in short periods of time. Given the significant parallels of working memory with other forms of long-term memory and its clear facility to be employed in experimental settings of short duration, working memory is a suitable candidate to identify early biomarkers of memory deficits ingeneral. In this study we assessed the cognitive performance and the electrophysiological response - via EEG signals - in a visual working memory recognition task that included the interference of past memories over the present ones. This 'proactive interference' effect is evaluated has a possible biomarker candidate for AD. Our findings reveal that subjects take longer reaction times in the recognition of visual items in the proactive interference condition in comparison to no interference. Additionally, we report an early (170-180 ms) and a later (430-450 ms) EEG components (ERP) that underlies the neural processing responsible for the resolution of this working memory interference. These two time intervals are interpreted as revealing the resolution of proactive interference at two difference stages of visual information processing ('letters'): the phonological (sub-lexical) and semantic (lexical) levels respectively. / --

This paper describes a new approach to the model base indexing stage of visual object recognition. Fast model base indexing of 3D objects is achieved by accessing a database of encoded 2D views of the objects using a fast 2D matching algorithm. The algorithm is specifically intended as a plausible solution for the problem of indexing into very large model bases that general purpose vision systems and robots will have to deal with in the future. Other properties that make the indexing algorithm attractive are that it can take advantage of most geometric and non-geometric properties of features without modification, and that it addresses the incremental model acquisition problem for 3D objects.

We investigate the differences --- conceptually and algorithmically --- between affine and projective frameworks for the tasks of visual recognition and reconstruction from perspective views. It is shown that an affine invariant exists between any view and a fixed view chosen as a reference view. This implies that for tasks for which a reference view can be chosen, such as in alignment schemes for visual recognition, projective invariants are not really necessary. We then use the affine invariant to derive new algebraic connections between perspective views. It is shown that three perspective views of an object are connected by certain algebraic functions of image coordinates alone (no structure or camera geometry needs to be involved).

Many real world datasets are characterized by having a long tailed distribution, with several samples for some classes and only a few samples for other classes. While many Deep Learning based solutions exist for object recognition when hundreds of samples are available, there are not many solutions for the case when there are only a few samples available per class. Recognition in the regime where the number of training samples available for each class are low, ranging from 1 to couple of tens of examples is called Lowshot Recognition. In this work, we attempt to solve this problem. Our framework is similar to [1]. We use a related dataset with sufficient number (a couple of hundred) of samples per class to learn representations using a Convolutional Neural Network (CNN). This CNN is used to extract features of the lowshot samples and learn a classifier . During representation learning, we enforce the learnt representations to obey certain property by using a custom loss function. We believe that when the lowshot sample obey this property the classification step becomes easier. We show that the proposed solution performs better than the softmax classifier by a good margin. / Master of Science / Deep learning, a branch of Artificial Intelligence(AI) is revolutionizing the way computers can learn and perform artificial intelligence tasks. The power of Deep Learning comes from being able to model very complex functions using huge amounts of data. For this reason, deep learning is criticized as being data hungry. Although AI systems are able to beat humans in many tasks, unlike humans, they still lack the ability to learn from less data. In this work, we address the problem of teaching AI systems with only a few examples, formally called the “low-shot learning”. We focus on low-shot visual recognition where the AI systems are taught to recognize different objects from images using very few examples. Solving the low-shot recognition problem will enable us to apply AI based methods to many real world tasks. Particularly in the cases where we cannot afford to collect huge number of images because it is either costly or it is impossible. We propose a novel technique to solve this problem. We show that our solution performs better at low-shot recognition than the regular image classification solution, the softmax classifier.

Geometric constructions using ruler and compass are being solved for thousands of years. Humans are capable of solving these problems without explicit knowledge of the analytical models of geometric primitives present in the scene. On the other hand, most methods for solving these problems on a computer require an analytical model. In this thesis, we introduce a method for solving geometrical constructions with access only to the image of the given geometric construction. The method utilizes Mask R-CNN, a convolutional neural network for detection and segmentation of objects in images and videos. Outputs of the Mask R-CNN are masks and bounding boxes with class labels of detected objects in the input image. In this work, we employ and adapt the Mask R- CNN architecture to solve geometric construction problems from image input. We create a process for computing geometric construction steps from masks obtained from Mask R- CNN and describe how to train the Mask R-CNN model to solve geometric construction problems. However, solving geometric problems this way is challenging, as we have to deal with object detection and construction ambiguity. There is possibly an infinite number of ways to solve a geometric construction problem. Furthermore, the method should be able to solve problems not seen during the...

This thesis is mostly about supervised visual recognition problems. Based on a general definition of categories, the contents are divided into two parts: one which models categories and one which is not category based. We are interested in data driven solutions for both kinds of problems. In the category-free part, we study novelty detection in temporal and spatial domains as a category-free recognition problem. Using data driven models, we demonstrate that based on a few reference exemplars, our methods are able to detect novelties in ego-motions of people, and changes in the static environments surrounding them. In the category level part, we study object recognition. We consider both object category classification and localization, and propose scalable data driven approaches for both problems. A mixture of parametric classifiers, initialized with a sophisticated clustering of the training data, is demonstrated to adapt to the data better than various baselines such as the same model initialized with less subtly designed procedures. A nonparametric large margin classifier is introduced and demonstrated to have a multitude of advantages in comparison to its competitors: better training and testing time costs, the ability to make use of indefinite/invariant and deformable similarity measures, and adaptive complexity are the main features of the proposed model. We also propose a rather realistic model of recognition problems, which quantifies the interplay between representations, classifiers, and recognition performances. Based on data-describing measures which are aggregates of pairwise similarities of the training data, our model characterizes and describes the distributions of training exemplars. The measures are shown to capture many aspects of the difficulty of categorization problems and correlate significantly to the observed recognition performances. Utilizing these measures, the model predicts the performance of particular classifiers on distributions similar to the training data. These predictions, when compared to the test performance of the classifiers on the test sets, are reasonably accurate. We discuss various aspects of visual recognition problems: what is the interplay between representations and classification tasks, how can different models better adapt to the training data, etc. We describe and analyze the aforementioned methods that are designed to tackle different visual recognition problems, but share one common characteristic: being data driven. / <p>QC 20140604</p>