Making scanned Arabic documents machine accessible using an ensemble of SVM classifiers

International Journal on Document Analysis and Recognition (IJDAR) - Raster-image PDF files originating from scanning or photographing paper documents are inaccessible to both text search engines...     

Salient Object Subitizing

We study the problem of salient object subitizing, i.e. predicting the existence and the number of salient objects in an image using holistic cues. This task is inspired by the ability of people to quickly and accurately identify the number of items within the subitizing range (1–4). To this end, we present a salient object subitizing image dataset of about 14 K everyday images which are annotated using an online crowdsourcing marketplace. We show that using an end-to-end trained convolutional neural network (CNN) model, we achieve prediction accuracy comparable to human performance in identifying images with zero or one salient object. For images with multiple salient objects, our model also provides significantly better than chance performance without requiring any localization process. Moreover, we propose a method to improve the training of the CNN subitizing model by leveraging synthetic images. In experiments, we demonstrate the accuracy and generalizability of our CNN subitizing model and its applications in salient object detection and image retrieval.     

Assessing remedial effectiveness through the blood lead:soil/dust lead relationship at the Bunker Hill Superfund Site in the Silver Valley of Idaho

The 21 square mile Bunker Hill Superfund Site in northern Idaho includes several thousand acres of contaminated hillsides and floodplain, a 365-acre abandoned lead/zinc smelter and is home to more than 7000 people in 5 residential communities. Childhood lead poisoning was epidemic in the 1970s with >75% of children exceeding 40 microg/dl blood lead. Health response activities have been ongoing for three decades. In 1991, a blood lead goal of 95% of children with levels less than 10 microg/dl was adopted. The cleanup strategy, based on biokinetic pathways models, was to reduce house dust lead exposure through elimination of soil-borne sources. An interim health intervention program, that included monitoring blood lead and exposures levels, was instituted to reduce exposures through parental education during the cleanup. In 1989 and 2001, 56% and 3% of children, respectively, exceeded the blood lead criteria. More than 4000 paired blood lead/environmental exposure observations were collected during this period. Several analyses of these data were accomplished. Slope factors derived for the relationship between blood lead, soil and dust concentrations are age-dependent and similar to literature reported values. Repeat measures analysis assessing year to year changes found that the remediation effort (without intervention) had approximately a 7.5 microg/dl effect in reducing a 2-year-old child's mean blood lead level over the course of the last ten years. Those receiving intervention had an additional 2-15 microg/dl decrease. Structural equations models indicate that from 40 to 50% of the blood lead absorbed from soils and dusts is through house dust with approximately 30% directly from community-wide soils and 30% from the home yard and immediate neighborhood. Both mean blood lead levels and percent of children to exceed 10 microg/dl have paralleled soil/dust lead intake rates estimated from the pathways model. Application of the IEUBK model for lead indicates that recommended USEPA default parameters overestimate mean blood lead levels, although the magnitude of over-prediction is diminished in recent years. Application of the site-specific model, using the soil and dust partitions suggested in the pathways model and an effective bioavailability of 18%, accurately predicts mean blood lead levels and percent of children to exceed 10 microg/dl throughout the 11-year cleanup period. This reduced response rate application of the IEUBK is consistent with the analysis used to originally develop the cleanup criteria and indicates the blood lead goal will be achieved.     

Piecemeal Learning of an Unknown Environment

We introduce a new learning problem: learning a graph bypiecemeal search, in which the learner must return every so often to its starting point (for refueling, say). We present two linear-time piecemeal-search algorithms for learningcity-block graphs: grid graphs with rectangular obstacles.     

Context effects in Stroop-like word and picture processing.

Presents a series of 6 experiments in which Stroop-like effects were generated by modally pure color–color, picture–picture, and word–word stimuli instead of the usual modally mixed color–word or picture–word stimuli. Naming, reading, and categorization tasks were applied. The Stroop inhibition was preserved with these stimuli but unexpectedly showed a semantic gradient only in the naming and not in the reading task. Word categorizing was slower and more interference prone than picture categorizing. These and other results can be captured by a model with two main assumptions: (a) semantic memory and the lexicon are separate, and (b) words have privileged access to the lexicon, whereas pictures and colors have privileged access to the semantic network. Such a model is developed and put to an initial test. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Using kernels for a video-based mouse-replacement interface

Some people cannot use their hands to control a computer mouse due to conditions such as cerebral palsy or multiple sclerosis. For these individuals, there are various mouse-replacement solutions. One approach is to enable them to control the mouse pointer using head motions captured with a web camera. One such system, the Camera Mouse, uses an optical flow approach to track a manually-selected small patch of the subject’s face, such as the nostril or the edge of the eyebrow. The optical flow tracker may lose the facial feature when the tracked image patch drifts away from the initially-selected feature or when a user makes a rapid head movement. To address the problem of feature loss, we developed and incorporated the Kernel-Subset-Tracker into the Camera Mouse. The Kernel-Subset-Tracker is an exemplar-based method that uses a training set of representative images to produce online templates for positional tracking. We designed the augmented Camera Mouse so that it can compute these templates in real time, employing kernel techniques traditionally used for classification. We propose three versions of the Kernel-Subset-Tracker, each using a different kernel, and compared their performance to the optical-flow tracker under five different experimental conditions. Our experiments with test subjects show that augmenting the Camera Mouse with the Kernel-Subset-Tracker improves communication bandwidth statistically significantly. Tracking of facial features was accurate, without feature drift, even during rapid head movements and extreme head orientations. We conclude by describing how the Camera Mouse augmented with the Kernel-Subset-Tracker enabled a stroke-victim with severe motion impairments to communicate via an on-screen keyboard.     

The influence of soil remediation on lead in house dust

Lead in house dust has long been recognized as a principal source of excess lead absorption among children at the Bunker Hill Superfund Site (BHSS) in northern Idaho. House dust lead concentration from homeowner's vacuum cleaner bags has been monitored since the epidemic of childhood lead poisoning in 1974. Geometric mean house dust lead concentrations decreased from >10000 mg/kg in 1974 to approximately 4000 mg/kg in 1975, in response to air pollution control initiatives at the defective primary lead smelter. After smelter closure, 1983 mean dust lead concentrations were near 3000 mg/kg and were most dependent on soil sources. Following emergency soil removals from public areas and roadsides and fugitive dust control efforts in the mid-1980s, house dust lead decreased by approximately 40-60% to 1200-1500 mg/kg. In 1992, a cleanup goal of 500 mg/kg dust lead community average, with no individual home exceeding 1000 mg/kg, was adopted. This goal was to be achieved by a combination of contaminated soil removals and fugitive dust control efforts throughout the 21 square mile BHSS. Continual reductions in house dust lead concentrations have been noted throughout the residential area soil cleanup. Geometric mean house dust lead concentrations averaged approximately 500-600 mg/kg from 1996 to 1999 and dropped below 500 mg/kg in 2000. Analysis of these data indicates that approximately 20% of the variance in dust lead concentrations is attributed to yard, neighborhood, and community soil lead concentrations. Since 1996, dust lead concentrations and dust and lead loading rates have also been measured by dust mats placed at entryways into the homes. Neighborhood soil lead concentrations, household hygiene, the number of adults living in the home, and the number of hours a child spends outdoors in summer explain approximately 26% of the variance in mat dust lead loading rates. It is estimated that post-remedial house dust lead concentrations will stabilize at 400-500 mg/kg, as compared to approximately 200 mg/kg in socio-economically similar background communities; the difference possibly attributed to residual soil concentrations (3-6 times background), recontamination of rights-of-way, tracking of non-residential mining district soils and dusts, fugitive dusts associated with the remediation, and residual structural or carpet dusts.