Datasets, features, learning, and models in visual recognition

Visual recognition is a fundamental research topic in computer vision. This dissertation explores datasets, features, learning, and models used for visual recognition. In order to train visual models and evaluate different recognition algorithms, this dissertation develops an approach to collect object image datasets on web pages using an analysis of text around the image and of image appearance. This method exploits established online knowledge resources (Wikipedia pages for text; Flickr and Caltech data sets for images). The resources provide rich text and object appearance information. This dissertation describes results on two datasets. The first is Berg’s collection of 10 animal categories; on this dataset, we significantly outperform previous approaches. On an additional set of 5 categories, experimental results show the effectiveness of the method. Images are represented as features for visual recognition. This dissertation introduces a text-based image feature and demonstrates that it consistently improves performance on hard object classification problems. The feature is built using an auxiliary dataset of images annotated with tags, downloaded from the Internet. Image tags are noisy. The method obtains the text features of an unannotated image from the tags of its k-nearest neighbors in this auxiliary collection. A visual classifier presented with an object viewed under novel circumstances (say, a new viewing direction) must rely on its visual examples. This text feature may not change, because the auxiliary dataset likely contains a similar picture. While the tags associated with images are noisy, they are more stable when appearance changes. The performance of this feature is tested using PASCAL VOC 2006 and 2007 datasets. This feature performs well; it consistently improves the performance of visual object classifiers, and is particularly effective when the training dataset is small. With more and more collected training data, computational cost becomes a bottleneck, especially when training sophisticated classifiers such as kernelized SVM. This dissertation proposes a fast training algorithm called Stochastic Intersection Kernel Machine (SIKMA). This proposed training method will be useful for many vision problems, as it can produce a kernel classifier that is more accurate than a linear classifier, and can be trained on tens of thousands of examples in two minutes. It processes training examples one by one in a sequence, so memory cost is no longer the bottleneck to process large scale datasets. This dissertation applies this approach to train classifiers of Flickr groups with many group training examples. The resulting Flickr group prediction scores can be used to measure image similarity between two images. Experimental results on the Corel dataset and a PASCAL VOC dataset show the learned Flickr features perform better on image matching, retrieval, and classification than conventional visual features. Visual models are usually trained to best separate positive and negative training examples. However, when recognizing a large number of object categories, there may not be enough training examples for most objects, due to the intrinsic long-tailed distribution of objects in the real world. This dissertation proposes an approach to use comparative object similarity. The key insight is that, given a set of object categories which are similar and a set of categories which are dissimilar, a good object model should respond more strongly to examples from similar categories than to examples from dissimilar categories. This dissertation develops a regularized kernel machine algorithm to use this category dependent similarity regularization. Experiments on hundreds of categories show that our method can make significant improvement for categories with few or even no positive examples.

Single particle motion of polymeric, colloidal, and biological materials

Small particles and their dynamics are of widespread interest due both to their unique properties and their ubiquity. Here, we investigate several classes of small particles: colloids, polymers, and liposomes. All these particles, due to their size on the order of microns, exhibit significant similarity in that they are large enough to be visualized in microscopes, but small enough to be significantly influenced by thermal (or Brownian) motion. Further, similar optical microscopy and experimental techniques are commonly employed to investigate all these particles. In this work, we develop single particle tracking techniques, which allow thorough characterization of individual particle dynamics, observing many behaviors which would be overlooked by methods which time or ensemble average. The various particle systems are also similar in that frequently, the signal-to-noise ratio represented a significant concern. In many cases, development of image analysis and particle tracking methods optimized to low signal-to-noise was critical to performing experimental observations. The simplest particles studied, in terms of their interaction potentials, were chemically homogeneous (though optically anisotropic) hard-sphere colloids. Using these spheres, we explored the comparatively underdeveloped conjunction of translation and rotation and particle hydrodynamics. Developing off this, the dynamics of clusters of spherical colloids were investigated, exploring how shape anisotropy influences the translation and rotation respectively. Transitioning away from uniform hard-sphere potentials, the interactions of amphiphilic colloidal particles were explored, observing the effects of hydrophilic and hydrophobic interactions upon pattern assembly and inter-particle dynamics. Interaction potentials were altered in a different fashion by working with suspensions of liposomes, which, while homogeneous, introduce the possibility of deformation. Even further degrees of freedom were introduced by observing the interaction of particles and then polymers within polymer suspensions or along lipid tubules. Throughout, while examination of the trajectories revealed that while by some measures, the averaged behaviors accorded with expectation, often closer examination made possible by single particle tracking revealed novel and unexpected phenomena.