Abstract: A system and method for automating and improving data extraction from a variety of document types, including both unstructured, structured, and nested content, is disclosed. The system and method incorporate an intelligent machine learning model that is designed to intelligently identify chunks of text, map the fields in the document, and extract multi-record values. The system is designed to operate with little to no human intervention, while offering significant gains in accuracy, data visualization, and efficiency. The architecture applies customized techniques including density-based adaptive text clustering, tabular data extraction based on hierarchical intelligent keyword searches, and natural language processing-based field value selection.

Abstract: Provided is a device for upscaling resolution based on a slice image, in which a low-resolution image is divided into a plurality of slice images so that a high-resolution image is generated. The device includes a convolution operation unit configured to convert a low-resolution input slice image into a high-resolution output slice image using a convolutional neural network. The convolutional neural network includes a cascading block configured to perform a convolution operation using a convolution filter having a predetermined size and a residual operation on an input feature map generated from the low-resolution input slice image to generate an output feature map, and an upscaling block configured to upscale the output feature map to generate the high-resolution output slice image.

Abstract: Methods and systems for high-resolution image manipulation are disclosed. An original high-resolution image to be manipulated is obtained, as well as a driving signal indicating a manipulation result. The original high-resolution image is down-sampled to obtain a low-resolution image to be manipulated. Using a trained manipulation generator, a low-resolution manipulated image and a motion field are generated from the low-resolution image. The motion field represent pixel displacements of the low-resolution image to obtain the manipulation indicated by the driving signal. A high-frequency residual image is computed from the original high-resolution image. A high-frequency manipulated residual image is generated using the motion field. A high-resolution manipulated image is outputted by combining the high-frequency manipulated residual image and a low-frequency manipulated image generated from the low-resolution manipulated image by up-sampling.

Abstract: A method includes receiving a first image. The method also includes detecting a plurality of edges in the first image. The method also includes connecting the edges. The method also includes identifying a contour in the first image based at least partially upon the connected edges. The method also includes determining a convex hull of the contour. The method also includes generating a second image comprising the convex hull.

Abstract: An image processing method for receiving M lens images and generating a projection image is disclosed. The method comprises: determining P optimal warping coefficients of P control regions in the projection image according to a 2D error table and the M lens images from an image capture module; generating M projection images according to the M lens images, a first vertex list and the P optimal warping coefficients; determining a seam for each of N seam regions; and, stitching two overlapping seam images to generate a stitched seam image for each seam region according to its corresponding seam. The 2D error table comprises multiple test warping coefficients and multiple accumulation pixel value differences in the P control regions. The P control regions are respectively located in the N seam regions respectively located in N overlap regions, where M>=2, N>=1 and P>=3.

Abstract: The invention provides a measurement method and system of undeformed chip thickness in micro-milling. This method includes the steps of: S1: acquiring a surface topography picture of the bottom of a flute after micro-milling; S2: extracting a tool mark at the central line of the flute from the surface topography picture; S3: calculating a spacing distance between adjacent tool marks and calculating the difference of equivalent cutting radius between adjacent cutter teeth based on the spacing distance between adjacent tool marks; and S4: reconstructing the instant undeformed chip thickness in micro-milling based on the difference of equivalent cutting radius between adjacent cutter teeth.

Abstract: The present disclosure provides a dimensional measurement method and device based on deep learning. The method includes capturing a target image of a target object, obtaining measurement data for the target image, and determining whether or not the target object is within a preset tolerance.

Abstract: A method of centering a correlation-based overlay includes resizing an overlay target image to a size smaller than an entire image size, defining first and second templates that are symmetrical to each other based on a diagonal in the resized image, and calculating a rough center coordinate by calculating a first correlation value representing a similarity symmetrical with respect to the diagonal between images of the first and second templates; defining first and second templates symmetrical based on a diagonal passing through the rough center coordinates in an original image of the overlay target image, calculating a fine center coordinate of the overlay target image by calculating a second correlation value representing a similarity symmetrical with respect to the diagonal between the images of the first and second templates; and centering an overlay key by moving a stage to a target position based on the fine center coordinates.

Abstract: A semantic segmentation architecture comprising an asymmetric encoder—decoder structure, wherein the architecture comprises further an adapter for linking different stages of the encoder and the decoder. The adapter amalgamates information from both the encoder and the decoder for preserving and refining information between multiple levels of the encoder and decoder. In this way the adapter aggregates features from different levels and intermediates between encoder and decoder.

Abstract: An artificial intelligence (AI) decoding apparatus includes a memory storing one or more instructions, and a processor configured to execute the stored one or more instructions to obtain image data corresponding to a first image that is downscaled from an original image by using first parameters of a first filter kernel included in a first deep neural network (DNN), reconstruct a second image corresponding to the first image, based on the obtained image data, and obtain a third image that is upscaled from the reconstructed second image, by performing an operation between the reconstructed second image and second parameters of a second filter kernel included in a second DNN corresponding to the first DNN. Each of the second parameters is represented by a product of a scale factor and one among integer values, and each of the integer values is 0 or ±2n, where n is an integer.

Abstract: An image processing apparatus for performing image quality processing on an image includes a feature extraction network and an image quality processing network including one or more modulation blocks, wherein each of the one or more modulation blocks includes a convolution layer, a modulation layer, and an activation layer for processing the image.

Abstract: A method for removing satellite trails from images includes the following: obtaining an initial image having stars, locating stars in the image, removing the located stars from the initial image to form a star-removed image, locating the satellite trail in the star-removed image, determining brightness information for the satellite trail in the initial image, creating a satellite-trail-containing image with only satellite trail brightness information based on the determined brightness information, and subtracting the satellite-trail-containing image from the initial image to remove the satellite trail so as to form a satellite-trail-removed image. Systems for removing satellite trails from images are also provided, as are other aspects.

Abstract: A system for producing a flipbook includes a processor that receives a video comprising a plurality of frames, selects a start frame and an end frame, and a plurality of frames therebetween. The processor can analyze the frames of the segment to determine an average rate of change of the plurality of frames and a threshold of relative image difference based on the average rate of change of the plurality of frames and a baseline frame rate. The processor can select, based on the results of its analysis, a plurality of selected frames, each of the selected frames being separated from two other selected frames by a sub-segment of the video, wherein each pair of adjacent frames comprises a relative image difference above the threshold and wherein each selected frame meets quality criteria not met by one or more local frames.

Abstract: A method for multi-camera self-supervised depth evaluation is described. The method includes training a self-supervised depth estimation model and an ego-motion estimation model according to a multi-camera photometric loss associated with a multi-camera rig of an ego vehicle. The method also includes generating a single-scale correction factor according to a depth map of each camera of the multi-camera rig during a time-step. The method further includes predicting a 360° point cloud of a scene surrounding the ego vehicle according to the self-supervised depth estimation model and the ego-motion estimation model. The method also includes scaling the 360° point cloud according to the single-scale correction factor to form an aligned 360° point cloud.

Abstract: An artificial intelligence (AI) decoding apparatus includes a memory storing one or more instructions, and a processor configured to execute the stored one or more instructions to obtain image data corresponding to a first image that is downscaled from an original image by using first parameters of a first filter kernel included in a first deep neural network (DNN), reconstruct a second image corresponding to the first image, based on the obtained image data, and obtain a third image that is upscaled from the reconstructed second image, by performing an operation between the reconstructed second image and second parameters of a second filter kernel included in a second DNN corresponding to the first DNN. Each of the second parameters is represented by a product of a scale factor and one among integer values, and each of the integer values is 0 or ±2n, where n is an integer.

Abstract: An image processor eliminates a character or object from a sequence of frames and then merges the resulting images with those of nearby frames, both preceding and succeeding, to synthesize the background of the sequence of frames.

Abstract: Systems and methods for analyzing image data to assess property damage are disclosed. According to certain aspects, a server may analyze segmented digital image data of a roof of a property using a convolutional neural network (CNN). The server may extract a set of features from a set of regions output by the CNN. Additionally, the server may analyze the set of features using an additional image model to generate a set of outputs indicative of a confidence level that actual hail damage is depicted in the set of regions.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating motion detection based on optical flow. One of the methods includes obtaining a first image of a scene in an environment taken by an agent at a first time point and a second image of the scene at a second later time point. A point cloud characterizing the scene in the environment is obtained. A predicted optical flow is determined between the first image and the second image. A respective initial flow prediction for the point that represents motion of the point between the two time points is determined. A respective ego motion flow estimate for the point that represents a motion of the point induced by ego motion of the agent is determined. A respective motion prediction that indicates whether the point was static or in motion between the two time points is determined.

Abstract: A method of processing an image of a scene including a road acquired by a vehicle-mounted camera to generate boundary data indicative of a boundary of an image region which represents an unoccupied area of the road, comprising: generating an LL sub-band image of an Nth level of an (N+1)-level discrete wavelet transform, DWT, decomposition of the image by iteratively low-pass filtering and down-sampling the image N times, where N is an integer equal to or greater than one; generating a sub-band image of an (N+1)th level by high-pass filtering the LL sub-band image of the Nth level, and down-sampling a result of the high-pass filtering, such that the sub-band image of the (N+1)th level has a pixel region having substantially equal pixel values representing the unoccupied area of the road in the image; and generating the boundary data by determining a boundary of the pixel region.

Abstract: An image restoration device obtains input data including input image information for each viewpoint, and generates an output image from warped image information generated by warping the input image information using a global transformation information of each viewpoint and disparity information of each viewpoint, using an image restoration model.