Abstract: The disclosure reduces an object shape estimation error even if an abnormality occurs in a device for obtaining a distance to the object. An information processing apparatus according to the disclosure obtains foreground information indicating regions corresponding to a foreground object in images captured by multiple image capturing devices that capture images of an image capturing region from varying directions, and distance information indicating a distance from a reference point to the foreground object, and obtains three-dimensional shape data indicating a foreground object three-dimensional shape based on the foreground information and the distance information. If the distance information is unobtainable by a first obtaining method, the information processing apparatus obtains the distance information by a second obtaining method, and at least one of the first and second obtaining methods uses at least part of the multiple image capturing devices as a stereo camera to obtain the distance information.

Abstract: The present invention provides a method for quantitatively rating the degree of similarity of images by using orientation, resizing of the digital image, pixilation of the digital image, creation of an image string suitable for generating a hit-score. In the generation of the image string color codes e.g. by the RGB system is used. The hit-score can be a percentage identity or homology between the image strings of the two digital images, said hit-score being the rating of the similarity between the two digital images. The invention also related to the use of the method in process of assessing insurance claims. Further the invention to the use of said method and for a computer program with instruction for carrying out the method.

Abstract: A system, computer readable medium and method trains a video transformer, using a machine learning engine, for human action recognition in a video. The method includes sampling video clips with varying temporal resolutions in global views and sampling the video clips from different spatiotemporal windows in local views. The machine learning engine is configured to match the global and local views in a framework of student-teacher networks to learn cross-view correspondence between local and global views, and to learn motion correspondence between varying temporal resolutions. The video transformer can output for display video clips in a manner that emphasizes attention to the recognized human action.

Abstract: A method for characterizing an intraoral organ. Modelling the organ as a digital three-dimensional model to be characterized (MTBC), including a mesh of points defining a surface. Placing the MTBC in a standardized configuration with respect to a digital three-dimensional initial reference model (IRM), including a mesh of points, called initial reference points (IRPs), the number of IRPs being less than 20% of the MTBC. Then, determining a final reference point (FRP), for each IRP, by a deformation algorithm. Then, determining a set of values determining the position of the FRP and/or a final elementary surface depending on the FRP, the algorithm determining the positions of the FRPs so that a final reference model consisting of a mesh of the final reference points matches the model to be characterized as closely as possible. Generating a characteristic vector grouping, in an ordered manner, the values determined for all the IRPs.

Abstract: A method for replacing a background in a picture, a device, a storage medium and a program product provided in the embodiments relate to image processing technologies and include: acquiring a basic background image of a preset scene, where the basic background image is an image obtained by shooting the preset scene in advance; acquiring a captured picture obtained by currently shooting the preset scene, and determining a current background image of the preset scene according to the basic background image and the captured picture; replacing the current background image in the captured picture with a preset background image to obtain a final captured picture.

Abstract: Disclosed are systems and methods for generating electronic instruments that implement digital transfers. The systems convert instruments to an electronic format using a digital imaging source that outputs image data. The image data is processed to determine content elements and segments of the electronic instrument and to extract transfer data. Neural networks implement artificial intelligence and machine learning technology that is used to secure the transfer instrument and detect inconsistencies or errors in the data.

Abstract: Techniques for animal identification based on animal behavior. A first video depicting a first animal associated with a user during a first visit to an enterprise location is received. One or more attributes pertaining to the first animal or to the user are received. A first behavior of the first animal as depicted in the first video is extracted and encoded into a first behavioral representation. One or more machine learning models are trained to identify the first animal based on the first behavioral representation and the one or more attributes.

Abstract: A method for multi-target multi-camera tracking includes: performing multi-target tracking on a video captured by each of multiple cameras, to extract a tracklet for each target appearing in the video, wherein multiple tracklets for multiple targets are extracted on the basis of multiple videos; extracting a feature for each tracklet; determining orientation of each tracklet based on orientation of the target relative to the camera; dividing the multiple tracklets into multiple groups based on the determined orientations; performing clustering on tracklets in each group based on the extracted features, such that tracklets corresponding to the same target are aggregated into an initial set; performing merging among initial sets obtained by performing clustering on respective groups, such that tracklets corresponding to the same target and having different orientations are merged into a final set; and using the tracklets in the final set as tracking information for the corresponding target.

Abstract: A high-precision odometry estimation method based on a double-layer filtering framework, comprising the following steps: acquiring sensing information of a binocular camera, a main body IMU, a legged auxiliary IMU, a legged joint motor and a foot end force sensor; fusing the rotation angle and rotational velocity of the legged joint motor, and the measurement information of the legged auxiliary IMU and the foot end force sensor by a bottom-layer legged filter to obtain a velocity observation in a legged auxiliary IMU coordinate system; then inputting the velocity observation into an upper-layer filter, and fusing the acquired observation information of the main body IMU and binocular camera by the upper-layer filter to complete information fusion and outputting an odometry estimation result of a robot. The method can provide an effective solution for the autonomous positioning of the quadruped mobile robot in the rapid movement.

Abstract: Provided are a method and a system for generating a diagnostic report. The method includes: obtaining raw scanning data of a target object, a reconstructed image of the target object, and clinical information of the target object; determining first lesion information of the target object and second lesion information of the target object based on the raw scanning data and the reconstructed image; determining whether the target object includes a lesion based on the first lesion information and the second lesion information; in response to determining that the target object includes the lesion, generating a target reconstructed image of the lesion using a simulation scanning model; obtaining third lesion information by performing feature extraction on the target reconstructed image; generating a structured diagnostic report using a report generation model based on the first lesion information, the second lesion information, the third lesion information, and the clinical information.

Abstract: Points are grouped from a point cloud using Concavity-induced Distance (CID). First, a set of seed points is sampled from the input point cloud. Then a grouping of points is conducted based on a nearest neighbor search computed using CID. These two steps enable two novel solutions in point cloud segmentation and scene abstraction. In one embodiment, CID is determined between two points residing within an object surface. In a second embodiment, a CID is determined between groups of points.

Abstract: Generating images includes generating encoded data by encoding input data into a latent space. The encoded data is decoded through a first decoder having first decoder layers by processing the encoded data through one or more of the first decoder layers. The encoded data is decoded through a second decoder having second decoder layers by processing the encoded data through one or more of the second decoder layers. An updated feature map is generated by replacing at least a portion of a feature map output from a selected layer of the first decoder layers with at least a portion of a feature map output from a selected layer of the second decoder layers. An image is generated by further decoding the updated feature map through one or more additional layers of the first decoder layers.

Abstract: A 3D medical image recognition method and apparatus, a device, a non-transitory computer-readable storage medium, and a computer program product are provided, which relate to the field of artificial intelligence. View rearrangement processing is performed in an ith-round feature extraction process on an (i?1)th-round 3D medical image feature to obtain 2D image features. The (i?1)th-round 3D medical image feature is obtained by performing (i?1)th-round feature extraction on a 3D medical image. Different 2D image features are features of the (i?1)th-round 3D medical image feature in different views. Semantic feature extraction processing is performed on each 2D image feature in different views. Feature fusion processing is performed on the image semantic features to obtain an ith-round 3D medical image feature. Image recognition processing is performed on an Ith-round 3D medical image feature obtained through Ith-round feature extraction to obtain image recognition of the 3D medical image.

Abstract: Systems and methods for super-resolution (SR) are described. The SR method includes determining a direction type of an input image based on a gradient of the input image, obtaining a first intermediate image based on the direction type using a look-up table (LUT) that stores an SR operation result of a pixel value combination corresponding to a kernel set mapped to the determined direction type, wherein the first intermediate image is an SR image corresponding to the input image, applying SR interpolation to the input image to obtain a baseline image, and obtaining an output image based on the first intermediate image and the baseline image.

Abstract: A computer-implemented method includes receiving a focal image stack. The focal image stack includes multiple images of a measurement object. Each of the images captures a region of a surface of the measurement object with a defined focal plane position in a depth direction. The defined focal plane positions of the plurality of images are different from each other. The method includes generating an initial image with an extended depth of field in the depth direction based on the focal image stack. The method includes generating a corrected image by correcting a set of imaging errors in the initial image. The set of imaging errors includes a distortion error.

Abstract: An image processing apparatus, according to one embodiment of the present invention, comprises: a receiver for receiving a first image and a second image; a first processor for outputting a third image by using at least one of the first image and the second image; a signal line bypassing at least one of the first image and the second image; a switch for connecting a signal output from the receiver to the first processor or the signal line; and a transmitter for transmitting at least two of the first to third images.

Abstract: A false video detection method is provided. The method includes: retrieving N video segments in a target video, wherein each video segment includes M image frames; performing a feature retrieving procedure on the N×M image frames to obtain feature maps corresponding to the N×M image frames; inputting the feature maps into a time-attention model to obtain concentrated feature; multiplying the concentrated feature with the feature maps to obtain first enhanced feature maps, adding the first enhanced feature maps with the feature maps to obtain second enhanced feature maps, and inputting the second enhanced feature maps into a fully connected layer to obtain a determination result outputted by the fully connected layer, wherein the determination result indicates that the target video is true or false.

Abstract: The present invention provides an image processing apparatus that inspects an image formed on a printing medium. The image processing apparatus reads an image formed on a printing medium and obtains a target image to be inspected, and corrects a position of control point among control points disposed on the target image, at or near an edge of the printing medium, and performs alignment using the corrected position of the control point so as to align the target image to a reference image using a non-rigid-body alignment, and inspects for a defect in the target image on a basis of the aligned target image and the reference image. In the correcting, a position of a control point at or near the edge is corrected on a basis of an interval between adjacent control points at or near the edge.

Abstract: A method of image correlation includes: obtaining a first image frame with a set of distinct features; obtaining a second image frame at least partially overlapping the first image frame; carrying out a cross-correlation operation of the first and second image frames to obtain a cross-correlation result indicating correlation values for different offset positions of the second image frame with respect to the first image frame, the highest correlation value with its associated offset position in the cross-correlation result indicating the most likely offset position of the second image frame with respect to the first image frame; and carrying out a reaggregation operation of the cross-correlation result to generate a distinct global maximum correlation value from a plurality of non-distinct correlation values in an aggregated cross-correlation result. The reaggregation operation involves carrying out a convolution of the cross-correlation result with a 1×2, 2×1, or 2×2 array of 1's.

Abstract: An image processing apparatus, according to one embodiment of the present invention, comprises: a receiver for receiving a first image and a second image; a first processor for outputting a third image by using at least one of the first image and the second image; a signal line bypassing at least one of the first image and the second image; a switch for connecting a signal output from the receiver to the first processor or the signal line; and a transmitter for transmitting at least two of the first to third images.