Abstract: The present disclosure provides a method and an apparatus of matching data, a device and a computer-readable storage medium, which are related to the field of artificial intelligence technology, and in particularly to the field of intelligent search and deep learning. The specific implementation solution includes: obtaining a first instance of a resume and a second instance of a job profile; determining, for a meta path, a resume feature data of the first instance and a profile feature data of the second instance, the meta path is a knowledge graph path from the resume to the job profile; and applying a classification model to the resume feature data of the first instance and the profile feature data of the second instance to determine a matching result between the first instance and the second instance.

Abstract: A method of training a model, a method of determining a word vector, a device, a medium, and a product are provided, which may be applied to fields of natural language processing, information processing, etc. The method includes: acquiring a first word vector set corresponding to a first word set; and generating a reduced-dimensional word vector for each word vector in the first word vector set based on a word embedding model, generating, for other word vector in the first word vector set, a first probability distribution in the first word vector set based on the reduced-dimensional word vector, and adjusting a parameter of the word embedding model so as to minimize a difference between the first probability distribution and a second probability distribution for the other word vector determined by a number of word vector in the first word vector set.

Abstract: A method, an apparatus, a device, a storage medium and a program product of performing a text matching are provided, which relate to a field of a computer technology, and in particular to natural language processing and deep learning technologies. The method includes: determining a word set and a plurality of semantic units from a text set, the word set is associated with a first predetermined attribute, and the text set contains a plurality of first texts indicating an object information and a plurality of second texts indicating an object demand information; generating a graph; and generating a final feature representation associated with the text set and the word set based on the graph and a graph convolution model, so as to perform the text matching.

Abstract: Techniques are described to combine image data from multiple images with different exposures into a relatively high dynamic range image. A first image of a scene may be generated with a first operational mode of an image processing system. A second image of the scene may be generated with a second different operational mode of the image processing system. The first image may be of a first spatial resolution, while the second image may be of a second spatial resolution. For example, the first spatial resolution may be higher than the second spatial resolution. The first image and the second image may be combined into an output image of the scene. The output image may be of a higher dynamic range than either of the first image and the second image and may be of a spatial resolution higher than the second spatial resolution.

Abstract: Techniques are described to combine image data from multiple images with different exposures into a relatively high dynamic range image. A first image of a scene may be generated with a first operational mode of an image processing system. A second image of the scene may be generated with a second different operational mode of the image processing system. The first image may be of a first spatial resolution, while the second image may be of a second spatial resolution. For example, the first spatial resolution may be higher than the second spatial resolution. The first image and the second image may be combined into an output image of the scene. The output image may be of a higher dynamic range than either of the first image and the second image and may be of a spatial resolution higher than the second spatial resolution.

Abstract: Techniques are described to combine image data from multiple images with different exposures into a relatively high dynamic range image. A first image of a scene may be generated with a first operational mode of an image processing system. A second image of the scene may be generated with a second different operational mode of the image processing system. The first image may be of a first spatial resolution, while the second image may be of a second spatial resolution. For example, the first spatial resolution may be higher than the second spatial resolution. The first image and the second image may be combined into an output image of the scene. The output image may be of a higher dynamic range than either of the first image and the second image and may be of a spatial resolution higher than the second spatial resolution.

Abstract: Techniques are provided to provide various pulse width modulation (PWM) schemes to embodiments of dual modulator display systems that may comprise a backlight of individually addressable and controllable light emitters. The backlight provides illumination to a light modulator for further conditioning of the light to be presented to a viewer. The backlight may be striped and each stripe is assigned a PWM scheme that effectively increases the bit depth of the controller for each stripe. The display system may allow a better matching of PWM periods to LCD frame rates to reduce visual artifacts. In another embodiment, the display system may detect a small bright feature to be rendered in the image data and, with a pre-assignment of light emitters to different partitions, the backlight controller may drive a subset of the light emitters according to the partitions.

Abstract: Techniques are described to combine image data from multiple images with different exposures into a relatively high dynamic range image. A first image of a scene may be generated with a first operational mode of an image processing system. A second image of the scene may be generated with a second different operational mode of the image processing system. The first image may be of a first spatial resolution, while the second image may be of a second spatial resolution. For example, the first spatial resolution may be higher than the second spatial resolution. The first image and the second image may be combined into an output image of the scene. The output image may be of a higher dynamic range than either of the first image and the second image and may be of a spatial resolution higher than the second spatial resolution.

Abstract: Techniques for performing panoramic scaling are disclosed that reduce visible distortion in a panoramic image. Further, techniques for performing combined YC adjustment and color conversion are disclosed that reduce the size and power requirements of video manipulation hardware by reducing the number of logic gates and memory buffers required when YC adjustment and color conversion are implemented as separate operations.

Abstract: Techniques for performing combined scaling, filtering, and/or scan conversion are disclosed that reduce the amount of line buffer space required in the overall design of a video processing system. In particular, coefficients from all or a sub-set of the scaling, filtering (smoothing/sharpening), and scan conversion filters are combined into one representative coefficient that can be applied in a single generic algorithm. Thus, implementation costs are reduced, particularly in a system-on-chip implementations.

Abstract: Techniques for performing panoramic scaling are disclosed that reduce visible distortion in a panoramic image. Further, techniques for performing combined YC adjustment and color conversion are disclosed that reduce the size and power requirements of video manipulation hardware by reducing the number of logic gates and memory buffers required when YC adjustment and color conversion are implemented as separate operations.

Abstract: Techniques for performing combined scaling, filtering, and/or scan conversion are disclosed that reduce that amount of line buffer space required in the overall design of a video processing system. In particular, coefficients from all or a sub-set of the scaling, filtering (smoothing/sharpening), and scan conversion filters are combined into one representative coefficient that can be applied in a single generic algorithm. Thus, implementation costs are reduced, particularly in system-on-chip implementations.

Abstract: A demosaic system and method that supports multiple CFA pattern inputs is disclosed. The demosaic system is capable of handling both RGB Bayer input and CMYG input and perform demosaic operations on both inputs to recover full-color images from the raw input images. The system uses a variable number gradient demosiac process. The process uses a 5×5 neighborhood of sensor pixel data centered at the pixel under consideration. The process calculates a set of gradients corresponding to different directions within the neighborhood of the sensor pixel data. A threshold value is determined and a subset of gradients is selected from the set of gradients that fall below the threshold value. The system calculates estimation values for the missing color value and the actual measured center pixel color value obtained from the sensor data on directions that are within the subset of gradients below the threshold.

Abstract: A demosaic system and method that supports multiple CFA pattern inputs is disclosed. The demosaic system is capable of handling both RGB Bayer input and CMYG input and perform demosaic operations on both inputs to recover full-color images from the raw input images. The system uses a variable number gradient demosiac process. The process uses a 5×5 neighborhood of sensor pixel data centered at the pixel under consideration. The process calculates a set of gradients corresponding to different directions within the neighborhood of the sensor pixel data. A threshold value is determined and a subset of gradients is selected from the set of gradients that fall below the threshold value. The system calculates estimation values for the missing color value and the actual measured center pixel color value obtained from the sensor data on directions that are within the subset of gradients below the threshold.

Abstract: A demosaic system and method that supports multiple CFA pattern inputs is disclosed. The demosaic system is capable of handling both RGB Bayer input and CMYG input and perform demosaic operations on both inputs to recover full-color images from the raw input images. The system uses a variable number gradient demosiac process. The process uses a 5×5 neighborhood of sensor pixel data centered at the pixel under consideration. The process calculates a set of gradients corresponding to different directions within the neighborhood of the sensor pixel data. A threshold value is determined and a subset of gradients is selected from the set of gradients that fall below the threshold value. The system calculates estimation values for the missing color value and the actual measured center pixel color value obtained from the sensor data on directions that are within the subset of gradients below the threshold.

Abstract: This invention relates generally to hardware for scaling and filtering video data and more specifically to algorithms and techniques for accelerating scaling and filtering operations on digital video data. The hardware is designed so that scaling and filtering operations are combined and performed simultaneously where possible to speed manipulation of the video data. Efficient design of the system allows memory buffers and logic gates to be shared or eliminated to reduce the size, cost and power requirements of the hardware implementation.

Abstract: A video processing system configured with DMA latency compensation is disclosed. This compensation helps minimize or otherwise mitigate shortages of data to the display, thereby improving the quality of displayed video. A relatively small line buffer is used to stage data for video processing. Should an underflow of data occur (where the buffer reading process is ahead of the buffer writing process), data is read from the previous line buffer. This not only prevents shortages of data to the display, but also provides data that is more likely to be relevant to the actual scene being displayed (as compared to random data).

Abstract: This invention relates generally to hardware for scaling and filtering video data and more specifically to algorithms and techniques for accelerating scaling and filtering operations on digital video data. The hardware is designed so that scaling and filtering operations are combined and performed simultaneously where possible to speed manipulation of the video data. Efficient design of the system allows memory buffers and logic gates to be shared or eliminated to reduce the size, cost and power requirements of the hardware implementation.

Abstract: Techniques for performing combined scaling, filtering, and/or scan conversion are disclosed that reduce the amount of line buffer space required in the overall design of a video processing system. In particular, coefficients from all or a sub-set of the scaling, filtering (smoothing/sharpening), and scan conversion filters are combined into one representative coefficient that can be applied in a single generic algorithm. Thus, implementation costs are reduced, particularly in a system-on-chip implementations.

Abstract: Techniques for performing panoramic scaling are disclosed that reduce visible distortion in a panoramic image. Further, techniques for performing combined YC adjustment and color conversion are disclosed that reduce the size and power requirements of video manipulation hardware by reducing the number of logic gates and memory buffers required when YC adjustment and color conversion are implemented as separate operations.