Abstract: Artistic styles extracted from one or more source images may be applied to one or more target images, e.g., in the form of stylized images and/or stylized video sequences. The extracted artistic style may be stored as a plurality of layers in a neural network, which neural network may be further optimized, e.g., via the fusion of various elements of the network's architectures. An optimized network architecture may be determined for each processing environment in which the network will be applied. The artistic style may be applied to the obtained images and/or video sequence of images using various optimization methods, such as the use of scalars to control the resolution of the unstylized and stylized images, temporal consistency constraints, as well as the use of dynamically adjustable or selectable versions of Deep Neural Networks (DNN) that are responsive to system performance parameters, such as available processing resources and thermal capacity.

Abstract: Supervised machine learning using neural networks is applied to denoising images rendered by MC path tracing. Specialization of neural networks may be achieved by using a modular design that allows reusing trained components in different networks and facilitates easy debugging and incremental building of complex structures. Specialization may also be achieved by using progressive neural networks. In some embodiments, training of a neural-network based denoiser may use importance sampling, where more challenging patches or patches including areas of particular interests within a training dataset are selected with higher probabilities than others. In some other embodiments, generative adversarial networks (GANs) may be used for training a machine-learning based denoiser as an alternative to using pre-defined loss functions.

Abstract: A computing device for performing image pre-processing for blending images receives a first image depicting a head. The computing device detects a portion of the head in the first image and calculates an image attribute of the portion in the first image. The computing device receives a second image and generates an adjusted second image by adjusting color pixels in the second image based on the calculated image attribute of the portion in the first image. The computing device blends the adjusted second image with the first image.

Abstract: The technology relates to navigating imagery that is organized into clusters based on common patterns exhibited when imagery is captured. For example, a set of captured images which satisfy a predetermined pattern may be determined. The images in the set of set of captured images may be grouped into one or more clusters according to the predetermined pattern. A request to display a first cluster of the one or more clusters may be received and, in response, a first captured image from the requested first cluster may be selected. The selected first captured image may then be displayed.

Abstract: A method and apparatus for detecting text are provided. The method includes: extracting features of a to-be-detected image; predicting using a character detection network a probability of each pixel point in the to-be-detected image being a character pixel point, and position information of each pixel point relative to a bounding box of a character including the pixel point when the pixel point is the character pixel point; determining position information of bounding boxes of candidate characters based on the prediction result of the character detection network; inputting the extracted features into a character map network, converting a feature map outputted by the character map network, and generating character vectors; determining a neighboring candidate character of each candidate character, and connecting each candidate character with an associated neighboring candidate character to form a character set; and determining a character area of the to-be-detected image.

Abstract: To accurately classify a form without using form layout information, the image processing apparatus utilizes a classifier that accepts a filled-in form whose image has been reduced into a specific size as an input and specifies the form type of the filled-in form. Machine learning has been performed to the classifier by a form type learning system and the form type learning system reduces an image of a filled-in form as an original document image, adds a noise to the original document image, which has not been reduced or has been reduced, to generate multiple images for machine learning, associates the form type of the original document image with the multiple images for machine learning as a label, and performs machine learning of the classifier using the multiple images for machine learning and the label as training data.

Abstract: A method for extracting a surface deformation feature of an object based on linear scanning three-dimensional point cloud is disclosed comprising: performing data acquisition by using a three-dimensional measurement sensor based on line structured light, and obtaining a three-dimensional point cloud data of a surface of the object after data pre-processing; eliminating influences of abnormal data and textures on extraction of sectional main profiles, to accurately obtain the sectional main profiles of the object; obtaining a binary image based on deformation feature points extracted from sections, in conjunction with a deformation feature knowledge base, and preliminarily positioning a deformation region based on sub-block images to obtain a set of target morphology sub-blocks; then performing a morphological operation on deformation feature points in the set of target morphology sub-blocks, and generating a deformation region of confidence ROC, and then performing a region growth to extract a target using ge

Abstract: The present disclosure provides a computer-implemented method and system for performing real time supervised detection of televised video ads in a media content of a broadcasted channel. The method includes reception of the media content and selection of a set of frames per second from the media content. The method includes extraction of keypoints from each selected frame and derivation of binary descriptors from extracted keypoints. The method includes assignment of weight value to each binary descriptor and creation of a special pyramid of the binary descriptors. The method includes obtaining a first vocabulary of binary descriptors for each selected frame and accessing a second vocabulary of binary descriptors. The method includes comparison of each binary descriptor in the first vocabulary with binary descriptors in second vocabulary and progressively scoring each selected frame of the media content. The method includes detection of a first ad in the media content.

Abstract: A system for automated geospatial image analysis comprising a deep learning model that receives orthorectified geospatial images, pre-labeled to demarcate objects of interest. The module presents marked geospatial images and a second set of unmarked, optimized, training geospatial images to a convolutional neural network. This process may be repeated so that an image analysis software module can detect multiple object types or categories. The image analysis software module receives orthorectified geospatial images from one or more geospatial image caches. Using a multi-scale sliding window submodule, image analysis software scans geospatial images, detects objects present and geospatially locates them.

Abstract: The present disclosure provides a computer-implemented method and system for adaptively reducing detection time in real time supervised detection of televised ads in media content of a channel. The method includes reception of the media content and selection of a set of frames per second from the media content. The method includes extraction of keypoints from each selected frame and derivation of binary descriptors from extracted keypoints. The method includes assignment of weight value to each binary descriptor and creation of a special pyramid of the binary descriptors. The method includes obtaining a first vocabulary of binary descriptors for each selected frame and accessing a second vocabulary of binary descriptors. The method includes comparison of each binary descriptor in the first vocabulary with binary descriptors in second vocabulary. The method includes progressively scoring each selected frame of the media content and detecting a first ad in the media content.

Abstract: The present disclosure provides a computer-implemented method and system for adaptively switching supervised detection strategy for watermarked and non-watermarked real time televised video advertisements. The televised video advertisements are present in a live stream of a media content of a broadcasted channel. The method includes selection of a set of frames per second from the media content. The method includes checking for one or more watermarked features in each selected frame of the media content. The method includes switching to a first detection strategy. The first detection strategy is associated with detection of a first ad in the live stream of the media content when the one or more watermarked feature are present in each checked frame in the selected set of frames.

Abstract: The present disclosure provides a computer-implemented method and system for hardware, channel, language and ad length agnostic detection of multi-lingual televised advertisements. The detection is performed across live streams of media content of one or more broadcasted channels. The method includes selection of a set of frames per second from a pre-defined set of frames. The method includes extraction of a pre-defined number of keypoints from each selected frame and derivation of a pre-defined number of binary descriptors from the extracted keypoints. The method includes creation of a special pyramid of the binary descriptors and accessing a second vocabulary of binary descriptors. The method includes comparison of each spatially identifiable binary descriptor from the first vocabulary with spatially identifiable binary descriptors in clusters of the second vocabulary. The method includes progressively scoring each selected frame and detection of the first ad in the live streams of the media content.

Abstract: An object recognition system is provided that includes a device configured to capture a video sequence formed from unlabeled testing video frames. The system includes a processor configured to pre-train a recognition engine formed from a reference set of CNNs on a still image domain that includes labeled training still image frames. The processor adapts the recognition engine to a video domain to form an adapted recognition engine, by applying a non-reference set of CNNs to a set of domains that include the still image and video domains and a degraded image domain. The degraded image domain includes labeled synthetically degraded versions of the labeled training still image frames included in the still image domain. The video domain includes random unlabeled training video frames. The processor recognizes, using the adapted engine, a set of objects in the video sequence. A display device displays the set of recognized objects.

Abstract: According to one embodiment, a monitoring system includes a monitoring terminal and a server. The monitoring terminal includes a detector, a tracking unit, a first selector, and a transmitter. The server includes a receiver, a second selector, a collation unit, and an output unit. The receiver receives a first best shot images from the monitoring terminal. The second selector performs second selection processing, as part of a predetermined selection processing other than a first selection processing, of selecting a second best shot image suitable for collation with a predetermined image from among the first best shot images. The collation unit performs collation processing of collating the second best shot image with the predetermined image. The output unit outputs a result of the collation processing.

Abstract: An image processing device includes a memory, and a processor coupled to the memory, wherein the processor is configured to execute acquiring a captured image including an original document region, detecting a color component value of a predetermined area in the captured image, detecting an edge in the predetermined area to acquire an edge amount indicating any one or both of a density of the edge and an edge intensity, and identifying, based on the color component value and the edge amount, a difference between a background of the original document region and a background of a background region obtained by removing the original document region from the captured image.

Abstract: A distance-based score approximation having improved computational efficiency is provided. Responsive to receiving a score request, a computing entity identifies an observation point based on a location indicated in the score request and defines a set of annuli comprising a plurality of concentric annuli centered on the observation point and defined by a predetermined maximum radius. The computing entity queries a geographic database for map information corresponding to geometry elements located within the predetermined maximum radius of the observation point and determines an intersection of each geometry element with each annulus. The computing entity determines a contribution for each intersection based at least in part on a size of the intersection, a measure assigned to the corresponding geometry element, and a representative radius of the corresponding annulus.

Abstract: The technology described herein recalibrates a structured light sensor in the field using time-of-flight sensor data. Structured light sensors are sensitive to mechanical changes that result in decreased accuracy. A structured light system calculates the range to an object by comparing a reference image to the actual image of the scene. The reference image is what the projected light pattern would look like on a flat object at a known distance. When the projected image changes, the reference image no longer matches the projected pattern. The calibration technology described herein captures a new reference image based on the current sensor characteristics using a time-of-flight capable sensor as the structured light imaging sensor.

Abstract: Artistic styles extracted from one or more source images may be applied to one or more target images, e.g., in the form of stylized images and/or stylized video sequences. The extracted artistic style may be stored as a plurality of layers in a neural network, which neural network may be further optimized, e.g., via the fusion of various elements of the network's architectures. An optimized network architecture may be determined for each processing environment in which the network will be applied. The artistic style may be applied to the obtained images and/or video sequence of images using various optimization methods, such as the use of scalars to control the resolution of the unstylized and stylized images, temporal consistency constraints, as well as the use of dynamically adjustable or selectable versions of Deep Neural Networks (DNN) that are responsive to system performance parameters, such as available processing resources and thermal capacity.

Abstract: An abnormality detection device includes: an estimator which estimates the amount of movement of a mobile body based on an image taken by a camera mounted on the mobile body; and a determiner which determines an abnormality in the camera by obtaining estimated information on the amount of movement of the mobile body as obtained in the estimator and actually observed information on the movement of the mobile body as detected by an external sensor, other than the camera, mounted on the mobile body.

Abstract: This disclosure describes a system for utilizing multiple image processing techniques to identify an item represented in an image. In some implementations, one or more image processing algorithms may be utilized to process a received image to generate item image information and compare the item image information with stored item image information to identify the item. When a similarity score identifying the similarity between the item image information and at least one of the stored item image information is returned, a determination may be made as to whether the similarity score is high enough to confidently identify the item. If it is determined that the similarity score is high enough to confidently identify the item, the other algorithms may be terminated and the determined identity of the item returned.