Abstract: A method may include receiving, from a camera, image data representing an object in an environment and determining, based on the image data, a vertical position within the image data of a bottom of the object. The method may also include determining an object bottom ratio between the vertical position and a height of the image data and determining, by way of a distance projection model and based on the object bottom ratio, an estimate of a physical distance between the camera and the object. The distance projection model may define, for each respective candidate object bottom ratio of a plurality of candidate object bottom ratios, a mapping between (i) the respective candidate object bottom ratio and (ii) a corresponding physical distance in the environment. The method may additionally include generating an indication of the estimate of the physical distance between the camera and the object.

Abstract: An observation device 10 includes an imaging device 17, storage 18, an acquirer 19, and a controller 20. The imaging device 17 generates an image. The storage 18 stores distance information. The distance information is a relationship between an image position and information pertaining to distance. The acquirer 19 acquires position information pertaining to the spatial position of a moving body. When the distance discrepancy between a measured distance and an estimated distance is at or above the first threshold, the controller 20 corrects the distance information using the measured distance. The measured distance corresponds to the position information. The estimated distance is obtained on the basis of the image position of the moving body in the image and the distance information.

Abstract: Embodiments of this application provide a target detection method and apparatus, an electronic device, and a computer storage medium. The method includes: obtaining a target image including a target object; performing instance segmentation on the target image, to obtain a segmentation mask corresponding to the target object; obtaining, based on the segmentation mask, position relationship features between target pixels in a target region in which the target object is located in the target image; obtaining position relationship features between standard pixels in a preset region of interest in a standard image, where the standard image includes a standard object corresponding to the target object; and matching the position relationship features between the target pixels and the position relationship features between the standard pixels, to obtain a correspondence between the target pixels and the standard pixels, and obtaining pose information of the target object based on the correspondence.

Abstract: Systems, apparatus, and methods for rendering content based on production element pose are disclosed. In an example, motion capture data of a production element is received, the production element moving from a first physical pose to a second physical pose. The motion capture data is processed to determine the coordinates of the second physical pose. A transformation of the second physical pose of the production element to a virtual pose of the production element is generated. A virtual model of the production element is updated, the virtual model comprising the virtual pose of the production element. The content is rendered based on the updated virtual model.

Abstract: Computer implemented methods and devices for determining dimensions or distances of head features are provided. The method includes identifying a plurality of features in an image of a head of a person. A real dimension of at least one target feature of the plurality of features or a real distance between at least one target feature of the plurality features and a camera device used for capturing the image is estimated based on probability distributions for real dimensions of at least one feature of the plurality of features and a pixel dimension of the at least one feature of the plurality of features.

Abstract: A method of detecting a reflection in a first thermal image captured by a thermal image sensor and, a second image is captured by a visible light sensor, near infrared sensor, or short-wave infrared sensor having a field of view which overlaps a field of view of the thermal image. A relationship between coordinates in the thermal image and the second image is determined. A first object is detected in a first position in the thermal image. A candidate image area in a second position in the second image is analysed to determine if an object equivalent to the first object is present, the second position corresponding to the first position according to the relationship between the coordinates in the thermal image and the second image. In response to a determination that there is no equivalent object in candidate image area, it is determined that first object is a reflection.

Abstract: Apparatuses, systems, and techniques are presented to detect one or more objects in one or more images. In at least one embodiment, one or more neural networks can be trained to detect one or more objects, in one or more unlabeled images, based at least in part upon one or more predicted segmentations of the one or more objects.

Abstract: A platform, device and process for capturing images of the surface of a gaming table and determining the quantity, identity, and arrangement of chips bet at a gaming table. Image data is captured corresponding to the one or more chips positioned in at least one betting area on a gaming surface of the respective gaming table and the data is processed to use a machine learning based approach to classify the chips using a combination of a depth information channel and at least one of a red channel, the green channel, and the blue channel.

Abstract: A method of masking or marking an object in an image stream is provided, including: generating one or more output image streams by processing an input image stream capturing a scene, including discarding pixel information about the scene provided by pixels of the input image stream, such that the discarded pixel information about the scene is not included in any output image stream; and detecting an object in the scene using the discarded pixel information, wherein generating the one or more output image streams includes masking or marking the detected object in at least one output image stream once deciding that the object is at least partially visible within this at least one output image stream. A corresponding device, computer program and computer program product are also provided.

Abstract: The invention includes a method for calibrating an object re-identification solution implementing an array of several cameras. For at least one pair of two cameras from among the several camera, the method includes detecting objects in images taken by each of the two cameras, and computing a normalized distance between a digital signature of each object detected by one of the two cameras, and that of each different object detected by the other one of the two cameras. The method also includes determining, as a function of said normalized distances, a distance threshold, called re-identification threshold, that will be used for re-identifying objects in the images taken by the two cameras. The invention also includes a computer program and a device implementing such a calibration method, and a method and a system for re-identifying individuals calibrated by such a calibration method.

Abstract: A transmission apparatus includes a communication interface configured to communicate with a reception apparatus, and a controller configured to acquire a depth image obtained by at least one depth camera measuring distances to a subject within a measurement range, identify, with reference to the acquired depth image, a window corresponding to a position of the subject from among a plurality of windows set within the measurement range and aligned in a depth direction, transmit a window identifier identifying the identified window to the reception apparatus via the communication interface, and for each pixel of the depth image, convert a measured distance into a relative distance within the window and transmit a resulting converted value as pixel data to the reception apparatus via the communication interface.

Abstract: System, processes and devices for monitoring betting activities using bet recognition devices and a server. Each bet recognition device has an imaging component for capturing image data for a gaming table surface. The bet recognition device receives calibration data for calibrating the bet recognition device. A server processor coupled to a data store processes the image data received from the bet recognition devices over the network to detect, for each betting area, a number of chips and a final bet value for the chips.

Abstract: A machine learning system in the present disclosure includes: an image pickup unit that photographs a product and acquires a product image; a preprocessing unit that generates an inspection target site image by clipping an image of an inspection target site of the product based on a setting file, and saves the generated inspection target site image in an image saving unit; and an inspection processing unit that performs a quality determination process on the inspection target site image of a quality determination object indicated as a quality determination target by production instruction information, in which when the inspection target site image saved in the image saving unit is relevant to a product that is designated by the production instruction information as a learning object that is not the quality determination target, the inspection target site image is accumulated in the image saving unit, as learning data.

Abstract: Techniques for identifying a 3D feature represented in images in a buffer are disclosed. The images comprise a first and second image, and the images provide data that is usable to determine a depth for the 3D feature. A 2D feature point that is an observation of the 3D feature is identified. A pseudo measurement is computed for the 2D feature point. This measurement includes a depth and an uncertainty value. A first joint uncertainty is determined for the depth using the pseudo measurement. The first image is ejected from the buffer, resulting in a reduction to an amount of the data that is usable to determine the depth. The pseudo measurement is tuned until a resulting second joint uncertainty, which is based on the pseudo measurement, is within a threshold level of similarity to the first joint uncertainty.

Abstract: A method and an apparatus for automatically recognizing an unmanned aerial vehicle, an electronic device, and a computer medium are provided.

Abstract: A method of autonomously monitoring a remote site, including the steps of locating a primary detector at a site to be monitored; creating one or more geospatial maps of the site using an overhead image of the site; calibrating the primary detector to the geospatial map using a detector-specific model; detecting an object in motion at the site; tracking the moving object on the geospatial map; and alerting a user to the presence of motion at the site. In addition thermal image data from a infrared cameras, rather than optical/visual image data, is used to create detector-specific models and geospatial maps in substantially the same way that optical cameras and optical image data would be used.

Abstract: In some examples, a system can access video data collected from one or more image sensors. The system can execute a human detection model to detect a person within a region of interest proximate to the environment based on the video data. The system can also execute an object detection model to detect a presence or absence of personal protection equipment (PPE) based on the video data. The system can associate the PPE with the person to detect an unsafe PPE status. The system can determine that the unsafe PPE status has persisted for a predetermined period of time. In response to determining that the unsafe PPE status has persisted for the predetermined period of time, the system can generate a signal indicating the unsafe PPE status.

Abstract: A method for estimating a porosity of a rock sample includes acquiring a calibrated digital image of the rock sample, extracting color and texture features from the digital image of the rock sample, and estimating the porosity of the rock sample using a model to evaluate the extracted color features and texture features. The model is configured to correlate digital image color and texture features with porosity.

Abstract: A mobile artificial neural network device is provided. The mobile artificial neural network device includes a camera configured to output a video of a product at a first frame rate, an AI recognition model configured to recognize a product information by receiving the product video, an artificial neural network processor configured to drive the AI recognition model at a second frame rate, and a display module configured to display the video of the product at the first frame rate and display the product information at the second frame rate.

Abstract: Provided are an apparatus and method for sensory substitution based on depth-time mapping. The apparatus for sensory substitution based on depth-time mapping according to the present invention includes an image segmentation unit configured to generate a plurality of depth images by segmenting three-dimensional (3D) image information based on depth, a sensory substitution unit configured to generate pieces of sensory substitution information by converting at least any one of the plurality of depth images into preset sensory information, a depth-time mapping unit configured to map the sensory substitution information to a time axis based on depth information corresponding to the depth image, and an output unit configured to sequentially output the pieces of sensory substitution information in order mapped to the time axis.