Abstract: According to one embodiment, an image processing device includes storage and a processor. The storage is configured to store a statistical model generated by learning bokeh that occurs in a first image affected by aberration of an optical system and varies non-linearly in accordance with a distance to a subject in the first image. The processor is configured to acquire a second image affected by the aberration of the optical system, perform color correction on the second image to reduce a number of colors expressed in the second image, and input a third image, obtained by performing the color correction on the second image, into the statistical model and acquire first distance information indicating a distance to a subject in the third image.

Abstract: An information processing apparatus (2000) includes a first analyzing unit (2020), a second analyzing unit (2040), and an estimating unit (2060). The first analyzing unit (2020) calculates a flow of a crowd in a capturing range of a fixed camera (10) using a first surveillance image (12). The second analyzing unit (2040) calculates a distribution of an attribute of objects in a capturing range of a moving camera (20) using a second surveillance image (22). The estimating unit (2060) estimates an attribute distribution for a range that is not included in the capturing range of the moving camera (20).

Abstract: A method for automatically reconstituting the architecture, along a reinforcing axis, of the reinforcement of a composite material, includes acquiring images of the reinforcement of the composite material, each image being acquired along a section plane perpendicular to the reinforcing axis; for each image acquired, detecting, using a neural network, barycentre and/or the circumference of each section of the reinforcing thread; for at least one acquired reference image, assigning a tag corresponding to a reinforcing thread, to each detected barycentre or circumference; for each other acquired image, assigning, to each detected barycentre and/or each detected circumference, the tag of the corresponding barycentre in the acquired reference image; reconstituting the architecture of each reinforcing thread from each detected barycentre and/or circumference having the tag of the reinforcing thread and the position on the reinforcing axis associated with the acquired image on which the barycentre and/or the circumf

Abstract: An image recognition method and system based on deep learning are provided. The image recognition system includes a first recognizing engine, at least one second recognizing engine and a processing circuit. The second recognizing engine is activated to recognize an image when the first recognizing engine is recognizing the image. The processing circuit determines whether to interrupt the first recognizing engine recognizing the image according to a result outputted by the second recognizing engine after the second recognizing engine completes recognition of the image.

Abstract: Provided is a robotic device, including: a chassis; a set of wheels; a control system; a battery; one or more sensors; a processor; a tangible, non-transitory, machine readable medium storing instructions that when executed by the processor effectuates operations including: capturing, with the one or more sensors, data of an environment of the robotic device and data indicative of movement of the robotic device; generating or updating, with the processor, a map of the environment based on at least a portion of the captured data; and generating or updating, with the processor, a movement path of the robotic device.

Abstract: An electronic device obtains an image that includes a face. The device performs feature extraction on the image, to obtain facial expression information corresponding to the face and facial feature information corresponding to the facial expression, wherein the facial feature information indicates an extent of the facial expression. The device determines facial emotion information according to the facial expression information. The device also determines facial feature expression information according to a target feature value corresponding to the facial emotion and the facial feature information. This expression recognition techniques disclosed herein can implement multi-task learning and reduce an amount of data required for model training, and can obtain both an emotion recognition result and a local expression recognition result, thereby improving efficiency and real-time performance of expression recognition and improving user experience.

Abstract: Provided is a tangible, non-transitory, machine readable medium storing instructions that when executed by the processor effectuates operations including: capturing visual readings to objects within an environment; capturing readings of wheel rotation; capturing readings of a driving surface; capturing distances to obstacles; determining displacement of the robotic device in two dimensions based on sensor readings of the driving surface; estimating, with the processor, a corrected position of the robotic device to replace a last known position of the robotic device; determining a most feasible element in an ensemble based on the visual readings; and determining a most feasible position of the robotic device as the corrected position based on the most feasible element in the ensemble and the visual readings.

Abstract: According to one embodiment, a defect classification apparatus includes a processor. The processor acquires a first design image which is an image based on design data created by design software and relates to a first inspection target, and acquires a first real image which is captured by imaging the first inspection target produced based on the design data. The processor converts the first design image to a reference image represented by using a second real image captured by imaging a second inspection target without a defect. The processor calculates a reliability of the reference image. The processor detects a defect in the first inspection target by comparing the reference image and the first real image and classifies the defect, based on the reliability.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training an image representation neural network.

Abstract: Explanation jobs may be performed for computer vision tasks. A request to execute an explanation job for a computer vision machine learning model may be received. The execution job may be performed, including extracting different features from the image, determining the respective relative importance values of the different features on inferences generated by the computer vision machine learning model. The result of the explanation job, including the generated heat maps may be provided.

Abstract: A meter readout on a meter has digits including a first digit, a second digit, etc. A sequence of images of the meter is obtained. The images include images of the digits in the meter readout. Automated recognition of the digits in the images result in likelihood arrays indicating the likelihoods for the digit values for the digits imaged in the meter images. Short chains of digit values are identified and spliced together to form a series of single digit, two-digit, three-digits, etc. paths that are built up based on the likelihood arrays. Various criteria are used to discard most of the chains and thereby avoid the combinatorial explosion of possible paths and thereby produce reliable meter readings without consuming considerable computational resources.

Abstract: Provided is a method for predicting attribute information, a coder, a decoder, and a storage medium. The coder determines a current Morton code corresponding to a point to be predicted in a point cloud to be coded, determines a target Morton code corresponding to the point to be predicted based on the current Morton code and according to a preset neighbor information table, judges whether a neighbor point of the point to be predicted exists in the point cloud to be coded according to the target Morton code, and performs prediction to obtain a predicted attribute value of the point to be predicted according to attribute reconstruction information of the neighbor point in response to that the neighbor point exists in the point cloud to be coded.

Abstract: The present disclosure relates to a camera device. The camera device and an electronic apparatus including the same according to an embodiment of the present disclosure include: a color camera; an IR camera; and a processor configured to separate a foreground region and a background region of an IR image from the IR camera, and to filter a color image from the color camera based on the separated foreground or background region, and generate a bokeh image for the color image. Accordingly, by using the IR image from the IR camera, the bokeh image for the color image from the color camera may be simply generated.

Abstract: The present disclosure relates to methods and devices for data or graphics processing including an apparatus, e.g., a GPU. The apparatus may receive at least one bitstream including a plurality of bits, each of the bits corresponding to a position in the at least one bitstream, and each of the bits being associated with color data. The apparatus may also arrange an order of the plurality of bits in the at least one bitstream, such that at least one of the bits corresponds to an updated position in the at least one bitstream. Additionally, the apparatus may convert, upon arranging the order of the bits, the color data associated with each of the plurality of bits in the at least one bitstream. The apparatus may also compress, upon converting the color data associated with each of the bits, the plurality of bits in the at least one bitstream.

Abstract: A set of residual elements useable to reconstruct a rendition of a first time sample of a signal is obtained. A set of spatio-temporal correlation elements associated with the first time sample is generated. The set of spatio-temporal correlation elements is indicative of an extent of spatial correlation between a plurality of residual elements and an extent of temporal correlation between first reference data based on the rendition and second reference data based on a rendition of a second time sample of the signal. The set of spatio-temporal correlation elements is used to generate output data.

Abstract: A method for image fusion may include obtaining a first image including a plurality of first pixels and a second image including a plurality of second pixels. Each of the plurality of second pixels may correspond to one of the plurality of first pixels. The method may further include classifying the plurality of first pixels into different categories which at least includes a first category and a second category. The method may further include fusing the first image with the second image based on a fusion operation between each of the plurality of first pixels and its corresponding second pixel to generate a fused image. The fusion operation associated with a first pixel belonging to the first category and the fusion operation associated with a first pixel belonging to the second category are performed according to different fusion rules.

Abstract: Described are systems and methods for direct sun imaging by a star tracker. Disclosed in a certain example is a direct sun imaging star tracker that includes an imaging sensor and a baffle. The baffle includes a star port, a sun port, and a beam splitter. The star port is configured to image first viewing environment while the sun port is configured to image a second viewing environment that includes the sun. The beam splitter is configured to combine electromagnetic radiation from the star port and the sun port into a combined image. In various examples, the systems and techniques described herein allow a star tracker to simultaneously view both the sun and the stars.

Abstract: Systems, methods, controllers, and techniques for addressing the parallax occlusion effect caused by non-collocated sensors are disclosed. A controller is configured to fuse image data received from an imaging device and depth data received from a depth sensor to form a mesh, project a ray from the imaging device to a pixel of the image data fused with a point of the depth data forming the mesh, determine an occlusion boundary surface within the depth data, and in response to determining that the ray intersects the occlusion boundary surface, determine that the imaging device is occluded from a fused point in the mesh.

Abstract: Embodiments discussed herein facilitate determination of one of a probability of prostate cancer recurrence-free survival or a risk factor associated with prostate cancer based on intra-tumor stromal morphology. Example embodiments can perform operations comprising: accessing a digitized histological image of a prostate of a patient, wherein the histological image comprises a region of interest associated with prostate cancer; identifying nuclei of intra-tumoral stromal cells within the region of interest; extracting, for the region of interest of the digitized histological image, one or more features describing the structure of the intra-tumoral stromal cells; and generating, via a model based at least on the one or more features, one of a probability of prostate cancer recurrence-free survival or a risk score associated with prostate cancer for the patient based at least on the extracted one or more features.

Abstract: Disclosed are methods, apparatus and systems for gesture recognition based on neural network processing. One exemplary method for identifying a gesture communicated by a subject includes receiving a plurality of images associated with the gesture, providing the plurality of images to a first 3-dimensional convolutional neural network (3D CNN) and a second 3D CNN, where the first 3D CNN is operable to produce motion information, where the second 3D CNN is operable to produce pose and color information, and where the first 3D CNN is operable to implement an optical flow algorithm to detect the gesture, fusing the motion information and the pose and color information to produce an identification of the gesture, and determining whether the identification corresponds to a singular gesture across the plurality of images using a recurrent neural network that comprises one or more long short-term memory units.