Andrew Rabinovich

Going deeper with convolutions

We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.

Object categorization using co-occurrence, location and appearance

In this work we introduce a novel approach to object categorization that incorporates two types of context-co-occurrence and relative location - with local appearance-based features. Our approach, named CoLA (for co-occurrence, location and appearance), uses a conditional random field (CRF) to maximize object label agreement according to both semantic and spatial relevance. We model relative location between objects using simple pairwise features. By vector quantizing this feature space, we learn a small set of prototypical spatial relationships directly from the data. We evaluate our results on two challenging datasets: PASCAL 2007 and MSRC. The results show that combining co-occurrence and spatial context improves accuracy in as many as half of the categories compared to using co-occurrence alone.

Weakly Supervised Object Localization with Stable Segmentations

Multiple Instance Learning (MIL) provides a framework for training a discriminative classifier from data with ambiguous labels. This framework is well suited for the task of learning object classifiers from weakly labeled image data, where only the presence of an object in an image is known, but not its location. Some recent work has explored the application of MIL algorithms to the tasks of image categorization and natural scene classification. In this paper we extend these ideas in a framework that uses MIL to recognize and localizeobjects in images. To achieve this we employ state of the art image descriptors and multiple stable segmentations. These components, combined with a powerful MIL algorithm, form our object recognition system called MILSS. We show highly competitive object categorization results on the Caltech dataset. To evaluate the performance of our algorithm further, we introduce the challenging Landmarks-18 dataset, a collection of photographs of famous landmarks from around the world. The results on this new dataset show the great potential of our proposed algorithm.

Advances in molecular labeling, high throughput imaging and machine intelligence portend powerful functional cellular biochemistry tools

Cellular behavior is complex. Successfully understanding systems at ever‐increasing complexity is fundamental to advances in modern science and unraveling the functional details of cellular behavior is no exception. We present a collection of prospectives to provide a glimpse of the techniques that will aid in collecting, managing and utilizing information on complex cellular processes via molecular imaging tools. These include: 1) visualizing intracellular protein activity with fluorescent markers, 2) high throughput (and automated) imaging of multilabeled cells in statistically significant numbers, and 3) machine intelligence to analyze subcellular image localization and pattern. Although not addressed here, the importance of combining cell‐image‐based information with detailed molecular structure and ligand‐receptor binding models cannot be overlooked. Advanced molecular imaging techniques have the potential to impact cellular diagnostics for cancer screening, clinical correlations of tissue molecular patterns for cancer biology, and cellular molecular interactions for accelerating drug discovery. The goal of finally understanding all cellular components and behaviors will be achieved by advances in both instrumentation engineering (software and hardware) and molecular biochemistry. J. Cell. Biochem. Suppl. 39: 194–210, 2002.

Model Order Selection and Cue Combination for Image Segmentation

Model order selection and cue combination are both difficult open problems in the area of clustering. In this work we build upon stability-based approaches to develop a new method for automatic model order selection and cue combination with applications to visual grouping. Novel features of our approach include the ability to detect multiple stable clusterings (instead of only one), a simpler means of calculating stability that does not require training a classifier, and a new characterization of the space of stabilities for a continuum of segmentations that provides for an efficient sampling scheme. Our contribution is a framework for visual grouping that frees the user from the hassles of parameter tuning and model order selection: the input is an image, the output is a shortlist of segmentations.

Scenes vs. objects: A comparative study of two approaches to context based recognition

Contextual models play a very important role in the task of object recognition. Over the years, two kinds of contextual models have emerged: models with contextual inference based on the statistical summary of the scene (we will refer to these as scene based context models, or SBC), and models representing the context in terms of relationships among objects in the image (object based context, or OBC). In designing object recognition systems, it is necessary to understand the theoretical and practical properties of such approaches. This work provides an analysis of these models and evaluates two of their representatives using the LabelMe dataset. We demonstrate a considerable margin of improvement using the OBC style approach.