Abstract: A system and method described herein is effective for generating automated visual analytics and player statistics for videos of sporting events. A dataset of videos is collected that comprises multiple teams. Training the networks on RGB and grayscale images affects the generalization ability of the network learned and augmenting more images using generative adversarial networks to the dataset helps further improves the performance.

Abstract: Apparatus configured to derive a set-selective syntax element indicating whether the predetermined block is to be predicted using one of a first set of intra-prediction modes comprising a DC intra prediction mode and angular prediction modes. If the set-selective syntax element indicates that the predetermined block is to be predicted using one of the first set of intra-prediction modes, the apparatus is configured to use a list of most probable intra-prediction modes. If the set-selective syntax element indicates that the predetermined block is not to be predicted using one of the first set of intra-prediction modes, the apparatus is configured to use a matrix-based intra-prediction mode. The list of most probable intra-prediction modes is formed such that the list of most probable intra-prediction modes is free of the DC intra prediction mode in case of the neighbouring blocks being predicted by any of the angular intra prediction modes.

Abstract: An image decoding method according to the present document comprises: decoding a current block on the basis of image information, wherein an MVP candidate list for the current block is derived on the basis of an inter prediction mode, which is derived on the basis of image information, and peripheral blocks of the current block; and deriving motion information of the current block on the basis of the MVP candidate list, wherein the peripheral blocks include a left bottom corner peripheral block, a left peripheral block, a right top corner peripheral block, a top peripheral block, and a left top corner peripheral block of the current block, the motion information includes an L0 motion vector for L0 prediction and/or an L1 motion vector for L1 prediction, the L0 motion vector is derived on the basis of an L0 motion vector predictor and an L0 motion vector difference, and the L1 motion vector is derived on the basis of an L1 motion vector predictor and an L1 motion vector difference.

Abstract: A method and a system for monitoring a spatial area in a personnel interlock are provided. Image data are acquired which represent a plurality of depth images of the spatial area which have been acquired simultaneously and from different positions in the area of the personnel interlock, in each case by a respective 3D image sensor arranged at the corresponding position. For each of the depth images, an associated characteristic depth image value is calculated. A total image value is obtained by combining the characteristic depth image values of the various depth images in accordance with a predetermined monotonic combination rule. An estimation result for the probability that more than one person is present in the spatial area is then determined.

Abstract: A vehicle display system according to an example embodiment of the present disclosure includes a camera configured to record images of an area outside a vehicle; a display configured to display a video feed of the recorded images to an occupant of the vehicle (e.g., on a display disposed within a vehicle cabin of the vehicle); and a processor operatively connected to the camera and display. The processor is configured to detect one or more objects in the recorded images; receive a selection of a particular one of the one or more objects from a vehicle occupant; and adjust the video feed based on the selection to keep the particular one of the one or more objects in a field of view of the video feed as the object moves relative to the vehicle.

Abstract: There is disclosed a computer-implemented method for lossy image or video compression, transmission and decoding, the method including the steps of: (i) receiving an input image at a first computer system; (ii) encoding the input image using a first trained neural network, using the first computer system, to produce a latent representation; (iii) quantizing the latent representation using the first computer system to produce a quantized latent; (iv) entropy encoding the quantized latent into a bitstream, using the first computer system; (v) transmitting the bitstream to a second computer system; (vi) the second computer system entropy decoding the bitstream to produce the quantized latent; (vii) the second computer system using a second trained neural network to produce an output image from the quantized latent, wherein the output image is an approximation of the input image. Related computer-implemented methods, systems, computer-implemented training methods and computer program products are disclosed.

Abstract: This application provides a method for constructing a motion information list in video encoding and decoding, an apparatus, and a device, relating to the field of video encoding and decoding technologies. The method includes: obtaining candidate motion information and a length of the motion information list; based on the length of the motion information list, determining target motion information to check for duplicates in the motion information list; checking the motion information list to for duplicates by comparing the target motion information with the motion information list; and updating the motion information list based on whether the motion information list has a duplicate of the target motion information.

Abstract: An example apparatus for encoding video frames includes a mask selector to select a subset of visual masks according to an actual target compression ratio and GOP configuration and a complexity estimator to estimate a picture level spatial/temporal complexity for a current frame. The example apparatus further includes a GOP adaptive visual mask selector to specify a visual mask from the subset of the visual masks corresponding to the estimated spatial and temporal complexity value a good enough picture QP deriver to derive a good enough picture QP value using the visual mask. The example apparatus also includes an adjustor to adjust the good enough picture QP value based on block level human visual system sensitivity and statistics of already encoded frames to obtain a final human visual system QP map.

Abstract: A computing device performs a method of decoding video data by receiving bitstream encoding a chroma block, a corresponding luma block, neighboring luma samples, and neighboring chroma samples; decoding the luma block, the plurality of neighboring luma samples, and the plurality of neighboring chroma samples; selecting a group of reference luma samples and a group of reference chroma samples; computing a threshold luma value from the plurality of reconstructed neighboring luma samples, and a threshold chroma value from the plurality of reconstructed neighboring chroma samples; determining a maximum luma value and a minimum luma value from the group of the reference luma samples; generating multi-model linear model (MMLM) including a first linear model between the minimum luma value and the threshold luma value, and a second linear model between the threshold luma value and the maximum luma value; and reconstructing the chroma block from the luma block using MMLM.

Abstract: An in-vehicle monitoring device includes a first imaging member, a second imaging member, and a controller. The first imaging member is configured to perform first imaging to capture a first image of one or more occupants in a passenger compartment of the vehicle. The second imaging member is configured to capture a second image of the one or more occupants. The controller is configured to execute a monitoring process on the first image and the second image. The first imaging member and the second imaging member are disposed in a center portion of the vehicle in a vehicle width direction of the vehicle so as to be arranged side-by-side in the vehicle width direction such that at least a portion of an imaging range of the first imaging member and at least a portion of an imaging range of the second imaging member overlap.

Abstract: Provided is a video coding method comprising that the width of a video picture is set to be a first multiple of a first number M, and the first number M is a positive integer. The height of the picture is set to be a second multiple of a second number N, and the second number N is a positive integer. The first number M and the second number N are predetermined. Additionally, it is determined that a merge sharing node is located outside of the picture. The merge sharing node is corresponding to a first region of the picture, a first set of coding units of the picture are within the first region, and the first set of coding units are coded with a merge mode. A constraint process is then applied such that the boundary of the merge sharing node is aligned with the boundary of the picture.

Abstract: Computer vision drilling systems and methods may be used with a drilling rig. The computer vision systems and methods may be used during drilling of a well to monitor the drilling equipment and personnel on the drilling site to provide safer drilling operations. The results from the computer vision drilling system may be used to cause corrective actions to be performed if a safety condition arises. In addition, computer vision systems and methods are provided to automatically monitor the drilling site and drilling operations, such as by tallying pipe in the drill string and by monitoring equipment for anomalous drilling conditions, and automatically taking corrective action as may be needed.

Abstract: The monitoring device of the disclosure includes a carrying unit having at least one carrying area for carrying a monitored object, at least part of the carrying area is provided with a corresponding image capturing unit and a corresponding optical path adjusting unit; the image capturing unit captures a monitoring image of the carrying area, each image capturing unit has a preset field of view, each image capturing unit forms the monitoring image according to light incident from the field of view; at least part of the carrying area is an extension area, the extension area is positioned outside the field of view of the image capturing unit; the optical path adjusting unit enables the light emitted from the extension area of the carrying area corresponding thereto to enter the image capturing unit from the field of view of the image capturing unit corresponding to the carrying area.

Abstract: A method for operating a field-of-vision display device for a motor vehicle, including an electrically controllable planar pixel assembly for generating a projection light beam having a display content and a reflection-suppressing deflection assembly, which is arranged on the planar pixel assembly and includes one or more flat reflection surfaces, extending along the planar pixel assembly at a predefined acute angle thereto and parallel to one another, for projecting the generated projection light beam onto a partially transparent, reflective projection pane, including a windshield of the motor vehicle, as a result of which a virtual display image shown in a field of vision of a user is generated behind the projection pane, wherein: the one or more reflection surfaces are each light-absorbing on the rear side thereof in order to suppress interfering reflection; at least one driving situation parameter is provided, in dependence on which different area segments of the planar pixel assembly are used to generate

Abstract: A full-automatic calibration method and apparatus oriented to a structured light 3D vision system are disclosed. For an erected 3D vision system, full-automatic calibration can be completed without moving the 3D vision system; and for an unfixed 3D vision system, full-automatic calibration can be completed without manual operation. By using the full-automatic calibration method oriented to the structured light 3D vision system, on one hand, non-professionals can easily complete calibration of structured light 3D imaging; and on the other hand, a problem of calibrating a large number of 3D cameras can also be solved.

Abstract: A multi-camera vision system is utilized onboard a work vehicle having an operator cabin and an apparatus associated with performance of a work task. The system includes: one or more display devices defining a plurality of display areas and a plurality of cameras carried by the work vehicle and coupled to display feeds at the display areas. A first camera of the cameras is oriented to selectively provide a first selected feed of at least a portion of the apparatus during performance of the work task. A controller is configured to display the first selected feed of the selected feeds with an insert, overlay, or icon that indicates the origin and orientation of the first camera as the first selected feed depicts performance of the work task by the apparatus.

Abstract: There is provided a work machine including a traveling body, a rotating platform mounted on the traveling body, an attachment mounted on the rotating platform, and including a work element, an imaging unit mounted on the rotating platform, and configured to capture a peripheral image, and an imaging controller configured to cause the imaging unit to capture the image, when an imaging range of the imaging unit includes a predetermined subject region designated in advance.

Abstract: There is disclosed a computer-implemented method for lossy image or video compression, transmission and decoding, the method including the steps of: (i) receiving an input image at a first computer system; (ii) encoding the input image using a first trained neural network, using the first computer system, to produce a latent representation; (iii) quantizing the latent representation using the first computer system to produce a quantized latent; (iv) entropy encoding the quantized latent into a bitstream, using the first computer system; (v) transmitting the bitstream to a second computer system; (vi) the second computer system entropy decoding the bitstream to produce the quantized latent; (vii) the second computer system using a second trained neural network to produce an output image from the quantized latent, wherein the output image is an approximation of the input image. Related computer-implemented methods, systems, computer-implemented training methods and computer program products are disclosed.

Abstract: An image sensor assembly includes at least one upconverter configured to detect light in a NIR waveband that is received from an object to be imaged and generate, based on the detected light, upconverted light that is outside of the NIR waveband; and at least one image sensor configured to detect the upconverted light.

Abstract: Video sources and inertial sensors are attached to a weapon and to goggles. A computer receives video images from the weapon- and goggles-mounted sources and inertial data from the sensors. The computer calculates a location for an image from the weapon-mounted source within an image from the goggles-mounted source using the inertial sensor data. The sensor-based location is checked (and possibly adjusted) based on a comparison of the images. A database contains information about real-world objects in a field of view of the goggles-mounted source, and is used to generate icons or other graphics concerning such objects.