Abstract: The present invention relates to a calculation device (10) for comparing dark-field X-ray images. The calculation device in (10) is configured for receiving a first dark-field X-ray image (11) describing first dark-field X-ray signals of a patient at an expiration state and for receiving a second dark-field X-ray image (12) describing second dark-field X-ray signals of the patient at an inspiration state. The calculation device is further (10) configured for normalizing the first dark-field X-ray signals of the first dark-field X-ray tin image (11) with a lung thickness value describing the lung thickness at the expiration state and for normalizing the second dark-field X-ray signals of the second dark-field X-ray image (12) with a lung thickness value describing the lung thickness at the inspiration state.

Abstract: One embodiment of the present invention sets forth a technique for extracting data from a document. The technique includes determining a first set of features associated with the document, wherein the first set of features comprises a set of region proposals that bound one or more portions of text within the document. The technique also includes applying a first machine learning model to the first set of features to generate a set of predictions associated with one or more key-value pairs included in the document. The technique further includes extracting the one or more key-value pairs from the document based on the set of predictions.

Abstract: A video enhancement method and apparatus, an electronic device, and a storage medium are described. The method comprises: extracting features from M frames of images, so as to obtain at least one first-scale image feature (S310); for each first-scale image feature, performing N-level down-sampling processing on the first-scale image feature, so as to obtain a second-scale image feature (S320); performing N-level up-sampling processing on the second-scale image feature, so as to obtain a third-scale image feature (S330), wherein the input of ith-level up-sampling processing is an image feature obtained after performing superimposition processing on the output of (N+1?i)th-level down-sampling processing and the output of (i?1)th-level up-sampling processing, and the multiple of jth-level up-sampling is the same as the multiple of (N+1?j)th-level down-sampling; and performing superimposition processing on the third-scale image feature and the first-scale image feature.

Abstract: Hierarchical systems and methods for automatic container type recognition from images are disclosed herein. An example embodiment includes a system for image analysis, comprising: a container recognition component; a character recognition component; and a 3D point cloud component; wherein the container recognition component is configured to receive an image and produce one of three outputs based on analysis of the image such that the output corresponds to either a container is identified, further analysis is performed by the character recognition component, or further analysis is performed by the 3D point cloud component.

Abstract: The present disclosure relates to security control methods and systems. The security control system may include at least one storage device storing a set of instructions; and one or more processors in communication with the at least one storage device, wherein when executing the set of instructions, the one or more processors are configured to direct the system to: obtain first data from a first device; obtain second data from a second device; associate and process the first data and/or the second data; and send the processed first data and second data to a server and/or a user terminal. The present disclosure can achieve linkage control of a plurality of access controls to meet the needs of users on indoor security.

Abstract: Computing platforms and methods are disclosed for verification of cleaning services to be performed in an area to be cleaned. Exemplary implementations may: identify, from among a plurality of cleaning regions relating to the area to be cleaned, a group of designated cleaning regions; randomly generate, with respect to the plurality of cleaning regions, a set of inspection regions; provide a user interface to prompt a user to capture an image of each inspection region; directly capture the images of each inspection region via the user interface; compare the captured images of each inspection region with corresponding reference images to determine whether each of the inspection regions has been properly cleaned; and output an inspection result based on the comparison of the captured images with the corresponding reference images. Implementations provide a method for virtually supervising otherwise unsupervised workers, which increases accountability and generates unprecedented visibility.

Abstract: High resolution object detection systems and methods provide accurate, real-time, one-stage processing, and include a backbone network configured to receive an input image and generate multi-scale feature representations, a feature fusion block configured to fuse the multi-sale feature representations, a plurality of representation transfer modules configured to isolate and decouple sub-task networks and the multi-scale feature representations, and a cascade refinement module configured to process each representation transfer module output to refine predictions. The backbone network generates a plurality of image features corresponding to each of a plurality of image scales and includes a plurality of convolutional layers and a stem block after the first convolutional layer, wherein the stem block improves feature extraction performance. The feature fusion block generates feature outputs for each of a plurality of image scales.

Abstract: Examples are disclosed herein that relate to detecting product requirements within a digitized document. One example provides a method comprising: identifying a first page as a summary page, the first page comprising a keyword that refers to a second page; and detecting in the first page a first instance of a pattern comprising a first text block adjacent to a first line. A first part name and a first requirement for a first part are extracted from the first text block. In the first page, a second instance of the pattern is detected comprising a second text block adjacent to a second line. The keyword and a second part name are extracted from the second text block. The second part name and a second requirement for a second part are extracted from the second page. The first requirement and the second requirement are output for storage in a data store.

Abstract: A method includes obtaining a first optical flow vector representing motion between consecutive video frames during a previous time step. The method also includes generating a first predicted optical flow vector from the first optical flow vector using a trained prediction model, where the first predicted optical flow vector represents predicted motion during a current time step. The method further includes refining the first predicted optical flow vector using a trained update model to generate a second optical flow vector representing motion during the current time step. The trained update model uses the first predicted optical flow vector, a video frame of the previous time step, and a video frame of the current time step to generate the second optical flow vector.

Abstract: A processor determines a cause of an image defect based on a test image that is obtained through an image reading process performed on an output sheet output from an image forming device. The processor generates an extraction image by extracting, from the test image, a noise point that is a type of image defect. Furthermore, the processor determines a cause of the noise point by using at least one of, in the extraction image: an edge strength of the noise point; a degree of flatness of the noise point; and a pixel value distribution of a transverse pixel sequence that is a pixel sequence traversing the noise point.

Abstract: A method includes capturing a first image associated with a portion of a display screen being shared. The method further includes rendering the first image in a preview window of the display screen being shared to form a second image. The second image is captured so as to determine whether the first image is duplicated in the second image. The duplication of the first image in the second image is masked to form a third image. The third image is rendered in the preview window.

Abstract: At least one embodiment relates to a method assigning a pixel value of an occupancy map either indicates that a depth value of at least one 3D sample of a point cloud frame projected along a same projection line is stored as a pixel value of at least one layer or equals a fixed-length codeword representing a depth value of at least one 3D sample projection along said projection line.

Abstract: Disclosed is a computer-implemented method for detecting one or more anatomic landmarks in medical image data. In an embodiment, the method includes receiving a medical image data set depicting a body part of a patient; and determining a first set of anatomic landmarks from a first representation of the medical image data set at a first resolution by applying a first trained function to the first representation of the medical image data set. Based on that, a second set of anatomic landmarks is determined from a second representation of the medical image data set at a second resolution, the second resolution being higher than the first resolution, by applying a second trained function different than the first trained function to the second representation of the medical image data set and using the first set of landmarks by the second trained function.

Abstract: The present disclosure relates to an object selection system that accurately detects and optionally automatically selects user-requested objects (e.g., query objects) in digital images. For example, the object selection system builds and utilizes an object selection pipeline to determine which object detection neural network to utilize to detect a query object based on analyzing the object class of a query object. In particular, the object selection system can identify both known object classes as well as objects corresponding to unknown object classes.

Abstract: A high-resolution image matching method and system are provided. The method includes performing regional fidelity down-sampling on an initial high-resolution image to obtain a multi-level low-resolution image, performing local matching on the obtained multi-level low-resolution images using a method with global probes to obtain a matching result of the low-resolution images, and performing reverse refinement on the obtained matching result of the low-resolution image using overall consistency of the image matching, to obtain the matching results of the high-resolution images at all levels until the matching results of the initial resolution images are obtained, so as to reduce the computational complexity of the image matching process and improve the accuracy of the matching result, and then the matching result of the high-resolution image is obtained through reverse refinement based on the overall consistency.

Abstract: A keypoint detection system includes: a camera system including at least one camera; and a processor and memory, the processor and memory being configured to: receive an image captured by the camera system; compute a plurality of keypoints in the image using a convolutional neural network including: a first layer implementing a first convolutional kernel; a second layer implementing a second convolutional kernel; an output layer; and a plurality of connections between the first layer and the second layer and between the second layer and the output layer, each of the connections having a corresponding weight stored in the memory; and output the plurality of keypoints of the image computed by the convolutional neural network.

Abstract: A fluorescence image registration method includes obtaining at least one fluorescence image of a biochip. An interior local area is selected. Sums of pixel values in the interior local area along a first direction and a second direction are obtained. A plurality of first template lines is selected to find a minimum total value of the sums of pixel values corresponding to the first template lines. Pixel-level correction is performed on a local area of the track line to obtain pixel-level track cross. Other track crosses on the biochip is obtained, and the pixel-level correction is performed on the other track crosses. The position of the pixel-level track line is corrected by a center-of-gravity method to obtain the subpixel-level position of the track line. The subpixel-level positions of all sites uniformly distributed on the biochip is obtained.

Abstract: A method for detecting motion information includes the following steps. First, a pixel array is provided for detecting an image of a measured object located in a first distance range or in a second distance range, and the pixel array includes a plurality of invisible image sensing pixels and a plurality of visible image sensing pixels. Then, image detection is conducted within the first distance range by using the invisible image sensing pixels to output a plurality of invisible images. Next, the image detection is conducted within the second distance range by using the visible image sensing pixels to output a plurality of visible images. Then, the plurality of invisible images and the plurality of visible images are analyzed by using a processing unit, so as to obtain motion information of the measured object. A pixel array for detecting motion information and an image sensor are also provided.

Abstract: An image processing device includes a storage device that previously stores a document image, a plurality of registered words, and a plurality of font characters, and a control device that functions as: a character region identifier that identifies a character region in the document image; an image acquirer that acquires an image of the character region; a text extractor that extracts a text from the image of the character region; a word identifier that identifies each of words in the text; a word determiner that determines whether each of the words is matched with one of the registered words; and a generator that generates a corrected text by replacing a target character of a non-matching word in the text with, among the font characters, a font character having a first degree of matching not lower than a first rate with the target character and a highest first degree of matching.

Abstract: This disclosure relates generally to system and method for parameter compression of capsule networks using deep features. The conventional capsule networks have distinct capability of retaining spatial correlations between extracted features but that comes at a cost of intensive computational, cost, memory usage and bandwidth requirement. The embodiments herein disclose a system and method for employing a lightweight deep features based capsule network that is capable of compressing the parameters. In an embodiment, the system includes a deep feature based capsule network such that the capsule layer is preceded by feature blocks. Said feature blocks comprises convolutional operation with a kernel size 3, followed by convolutional operation with kernel of size 1, and a Batch Normalization layer, and hence are able to extract deep features.