Abstract: Systems, methods, and computer program products implementing techniques for training classifiers. The techniques include receiving a training set that includes positive samples and negative samples, receiving a restricted set of linear operators, and using a boosting process to train a classifier to discriminate between the positive and negative samples. The boosting process is an iterative process. The iterations include a first iteration where a classifier is trained by (1) testing some, but not all linear operators in the restricted set against a weighted version of the training set, (2) selecting for use by the classifier the linear operator with the lowest error rate, and (3) generating a re-weighted version of the training set. The iterations also include subsequent iterations during which another classifier is trained by repeating steps (1), (2), and (3), but using in step (1) the re-weighted version of the training set generated during a previous iteration.

Abstract: Systems, methods, and computer program products, implementing techniques for detecting objects using a soft cascade. The techniques include receiving a digital data segment and determining whether the digital data segment resembles an object of interest by passing the digital data segment through a cascade. The cascade includes an ordered sequence of stages and a rejection function after each stage that determines whether to reject the digital data segment at the current stage as not resembling the object of interest, or to allow the digital data segment to pass to the next stage of evaluation. The rejection function allows the digital data segment to fail the current stage and still pass to the next stage.

Abstract: Methods, systems, and computer program products used to locate a feature in a digital image. A first search is performed in the image to find candidate faces, where each candidate face found is a group of pixels in the image that satisfies a first pattern-matching criterion. A second search is performed in the image to find candidate eyes, where each candidate eye found is a group of pixels in the image that satisfies a second pattern-matching criterion. A third search is performed within each candidate face that includes at least one overlapping candidate eye to find red pupils, where each red pupil found is a group of pixels in the image. A color modification process is applied to the red pupils.

Abstract: A conveyor cleaning system includes a treating region, a processing region, and a conveyor belt traveling from the treating region to the processing region. A distributor positioned across the treating region deposits a film of ozonated water on the conveyor belt. In the processing region, the conveyor belt is exposed to soiling agents such as grease and other animal matter. A high-pressure rinser may be positioned over the treating section such that the distributor is positioned between the high-pressure rinser and the processing region. Ozonated water may be generated by a system including first and second tanks. The first tank is filled with ozone gas and water. After a predetermined time, undissolved ozone gas is released. The contents of the first tank are then transferred to a second tank along with more ozone gas. After waiting a predetermined time period, undissolved ozone gas is released from the second tank.

Abstract: A system maintains data from different resolution levels of an image in textures of a graphics processing unit (GPU). Image data is organized into multiple resolution layers of an image. Data from the lower resolution level(s) is used to process an image while higher resolution data is loaded. In one embodiment, a first resolution representation of the image having a lowest resolution level using data resident in the GPU memory is drawn prior to drawing a portion of a second resolution representation having a higher resolution level using data resident in the GPU memory.

Abstract: Methods and apparatus, including computer program products, implementing techniques for training an attentional cascade. An attentional cascade is an ordered sequence of detector functions, where the detector functions are functions that examine a target image and return a positive result if the target image resembles an object of interest and a negative result if the target image does not resemble the object of interest. A positive result from one detector function leads to consideration of the target image by the next detector function, and a negative result from any detector function leads to rejection of the target image. The techniques include training each detector function in the attentional cascade in sequence starting with the first detector function. Training a detector function includes selecting a counter-example set. Selecting a counter-example set includes selecting only images that are at least a minimum difference from an example set.

Abstract: A method, system, and computer-readable storage medium are disclosed for determining one or more symmetries in an image comprising a plurality of pixels. A symmetry value may be automatically determined for each of the plurality of pixels. The symmetry value may indicate the strength of one or more symmetries in the image for the respective pixel. The symmetry value may be stored for each of the pixels.

Abstract: Systems, methods, and computer program products implementing techniques for training classifiers. The techniques include receiving a training set that includes positive images and negative images, receiving a restricted set of linear operators, and using a boosting process to train a classifier to discriminate between the positive and negative images. The boosting process is an iterative process. The iterations include a first iteration where a classifier is trained by (1) testing some, but not all linear operators in the restricted set against a weighted version of the training set, (2) selecting for use by the classifier the linear operator with the lowest error rate, and (3) generating a re-weighted version of the training set. The iterations also include subsequent iterations during which another classifier is trained by repeating steps (1), (2), and (3), but using in step (1) the re-weighted version of the training set generated during a previous iteration.

Abstract: Methods, systems, and computer program products used to locate a feature in an image, including identifying one or more candidate features in an image, where each candidate feature is a group of pixels in the image that satisfies a pattern-matching criterion. A best candidate feature is selected from the one or more candidate features, and a parameterized shape is fit to the image in the region of the best candidate feature to compute a feature shape corresponding to the best candidate feature. Particular implantations can include one or more of the following features. The candidate feature is a candidate pupil and the feature shape is an ellipse. Fitting the parameterizes shape to the mage includes applying an iterative process varying shape parameters.

Abstract: A conveyor cleaning system includes a treating region, a processing region, and a conveyor belt traveling from the treating region to the processing region. A distributor positioned across the treating region deposits a film of ozonated water on the conveyor belt. In the processing region, the conveyor belt is exposed to soiling agents such as grease and other animal matter. A high-pressure rinser may be positioned over the treating section such that the distributor is positioned between the high-pressure rinser and the processing region. Ozonated water may be generated by a system including first and second tanks. The first tank is filled with ozone gas and water. After a predetermined time, undissolved ozone gas is released. The contents of the first tank are then transferred to a second tank along with more ozone gas. After waiting a predetermined time period, undissolved ozone gas is released from the second tank.

Abstract: A conveyor cleaning system includes a treating region, a processing region, and a conveyor belt traveling from the treating region to the processing region. A distributor positioned across the treating region deposits a film of ozonated water on the conveyor belt. In the processing region, the conveyor belt is exposed to soiling agents such as grease and other animal matter. A high-pressure rinser may be positioned over the treating section such that the distributor is positioned between the high-pressure rinser and the processing region. Ozonated water may be generated by a system including first and second tanks. The first tank is filled with ozone gas and water. After a predetermined time, undissolved ozone gas is released. The contents of the first tank are then transferred to a second tank along with more ozone gas. After waiting a predetermined time period, undissolved ozone gas is released from the second tank.

Abstract: A high pressure water stream (14) is discharged onto a surface to be cleaned. An ozone/water stream (16) is discharged on the same surface for sanitizing the surface. The high pressure water and ozone/water streams (14, 16) are discharged simultaneously along closely adjacent paths that are either parallel (FIG. 3) or concentric (FIG. 2). The water pressure is at least about 100 p.s.i. and is preferably between 100 p.s.i. and 1000 p.s.i. The nozzles that discharge the streams (14, 16) may be movable relative to the object(s) that receives the high pressure water and ozone/water (FIG. 1) Or, they may be fixed and the object my be movable relative to them (FIG.

Abstract: Methods, systems, and computer program products used to locate a feature in a digital image. A first search is performed in the image to find candidate faces, where each candidate face found is a group of pixels in the image that satisfies a first pattern-matching criterion. A second search is performed in the image to find candidate eyes, where each candidate eye found is a group of pixels in the image that satisfies a second pattern-matching criterion. A third search is performed within each candidate face that includes at least one overlapping candidate eye to find red pupils, where each red pupil found is a group of pixels in the image. A color modification process is applied to the red pupils.

Abstract: A system and method for high-speed execution of graphics application programs, including shading language instructions, that utilize 3D graphics hardware. The method involves expressing a graphics computation in a platform-independent procedural shading expression, converting the expression (i.e., user application program) into an intermediate representation such as a tree, and then translating it into a sequence of parametric shading expressions. The method can alternatively processes the intermediate tree representation using a cost-based, tree-grammar recognizer such as iburg. The result is a platform-specific and least-cost, in terms of the underlying hardware, sequence of parametric shading expressions that realizes the graphics computation. The system and method is useful in translating platform-independent RenderMan™ Shading Language programs into fast-executing, platform-specific OpenGL™ or Direct3D™ executable code.

Abstract: An apparatus and method for linking a video and an animation. A data structure containing elements that correspond to respective frames of a first video is generated. Information that indicates an image in an animation that has been created from a second video is stored in one or more of the elements of the data structure.