Abstract: Systems and methods for generating simulation data based on real-world dynamic objects are provided. A method includes obtaining two- and three-dimensional data descriptive of a dynamic object in the real world. The two- and three-dimensional information can be provided as an input to a machine-learned model to receive object model parameters descriptive of a pose and shape modification with respect to a three-dimensional template object model. The parameters can represent a three-dimensional dynamic object model indicative of an object pose and an object shape for the dynamic object. The method can be repeated on sequential two- and three-dimensional information to generate a sequence of object model parameters over time. Portions of a sequence of parameters can be stored as simulation data descriptive of a simulated trajectory of a unique dynamic object. The parameters can be evaluated by an objective function to refine the parameters and train the machine-learned model.

Abstract: In order to improve early emergence stages discrimination of plants, an early emergence app is proposed based on a neural network optionally with ‘attention’ mechanisms, which detects the number of plants that opened after sowing and are present in the field. This way the farmer can easily determine, if crop density targets are met at an early stage after sowing and optionally receive recommendations on catch crop.

Abstract: A method of generating a segmentation outcome which indicates individual instances of one or more object classes for an image in a sequence of images is disclosed.

Abstract: An image processing apparatus includes a photographing device, an image processor and a fault diagnostic device. The photographing device outputs image data. The image processor includes an electronic circuit having recognition algorithms respectively for image recognition of a predetermined recognition target included in the image data. The fault diagnostic device determines that a fault occurs in the image processor, based on a condition that a first recognition result of recognizing test image data including a characteristic pattern also included in the recognition target in case of the image processor receiving the test image data is different from an expected value as a second recognition result of recognizing the test image data in case of one of the recognition algorithms receiving the test image data.

Abstract: Embodiments of the present disclosure relate to an image processing method, a device and a computer-readable storage medium. The method includes obtaining a first reference image and a second reference image of a reference object. The method further includes determining a first constraint and a second constraint based on the first reference image and the second reference image. The first constraint is related to a difference between different depth maps of the same reference images, and the second constraint is related to a depth consistency of different reference images. The method further includes training a depth estimation model at least based on the first constraint and the second constraint. In this way, an obtained trained depth estimation model may provide more accurate depth information.

Abstract: Systems and methods are disclosed for predicting one or more medical conditions utilizing digital images and employing artificial intelligent algorithms. The system offers accurate predictions utilizing quantized pre-trained deep learning model. The pre-trained deep learning model is trained on data samples and later refined as the system processes more digital images or new medical conditions are incorporated. One pre-trained deep learning model is used to predict the probability of one or more medical conditions and identify locations in the digital image effected by the one or more medical conditions. Further, one pre-trained deep learning model utilizing additional data and plurality of digital images, forecasts rate of infection and spread of the medical condition over time.

Abstract: An object attribute inference method, a storage medium and an electronic device. The method includes the following steps: training a preset neural network by using a training sample comprising an action sequence and a target object attribute tag; where the action sequence is an interactive action sequence of a human body and an object. In this way, the object attribute inference model obtained through training can recognize the object attributes from the interaction action of the human body and the object, the object attributes include and are not limited to the weight, the shape, the hardness degree and the like of the object, and therefore the inference module obtained through training has the advantages of being universal in object attribute inference and wide in application range.

Abstract: A method to measure a cognitive load based upon ocular information of a subject includes the steps of: providing a video camera configured to record a close-up view of at least one eye of the subject; providing a computing device electronically connected to the video camera and the electronic display; recording, via the video camera, the ocular information of the at least one eye of the subject; processing, via the computing device, the ocular information to identify changes in ocular signals of the subject through the use of convolutional neural networks; evaluating, via the computing device, the changes in ocular signals from the convolutional neural networks by a machine learning algorithm; determining, via the machine learning algorithm, the cognitive load for the subject; and displaying, to the subject and/or to a supervisor, the cognitive load for the subject.

Abstract: An interaction detection method is provided. The method includes the steps of: acquiring one or more frame images; acquiring pose data of a first object using information on a plurality of feature points detected for the first object from a first frame image; determining occurrence of an interaction of the first object using the pose data of the first object; estimating a region of interest (ROI) of the first object using the information on the plurality of feature points; and acquiring information on a product corresponding to the region of interest of the first object.

Abstract: The technology disclosed relates to automatically (e.g., programmatically) initializing predictive information for tracking a complex control object (e.g., hand, hand and tool combination, robot end effector) based upon information about characteristics of the object determined from sets of collected observed information. Automated initialization techniques obviate the need for special and often bizarre start-up rituals (place your hands on the screen at the places indicated during a full moon, and so forth) required by conventional techniques. In implementations, systems can refine initial predictive information to reflect an observed condition based on comparison of the observed with an analysis of sets of collected observed information.

Abstract: Systems and methods are provided for synthesizing information from multiple image series of different kernels into a single image series, and also for converting a single baseline image series of a kernel reconstructed by a CT scanner to image series of various other kernels, using deep-learning based methods. For multi-kernel synthesis, a single set of images with desired high spatial resolution and low image noise can be synthesized from multiple image series of different kernels. The synthesized kernel is sufficient for a wide variety of clinical tasks, even in circumstances that would otherwise require many separate image sets. Kernel conversion may be configured to generate images with arbitrary reconstruction kernels from a single baseline kernel. This would reduce the burden on the CT scanner and the archival system, and greatly simplify the clinical workflow.

Abstract: A method of performing scene-dependent lens shading correction (SD-LSC) is provided. The method includes collecting scene information from a Bayer thumbnail of an input image; generating a standard red green blue (sRGB) thumbnail by processing the Bayer thumbnail of the input image to simulate white balance (WB) and pre-gamma blocks; determining a representative color channel ratio of the input image based on the scene information and the sRGB thumbnail; determining an ideal grid gain of the input image based on the representative color channel ratio and a grid gain of the input image; merging the ideal grid gain and the grid gain of the input image to generate a new grid gain; and applying the new grid gain to the input image.

Abstract: Described herein are systems, methods, and instrumentalities associated with denoising medical images such as fluoroscopic images using deep learning techniques. A first artificial neural network (ANN) is trained to denoise an input medical image in accordance with a provided target noise level. The training of the first ANN is conducted by pairing a noisy input image with target denoised images that include different levels of noise. These target denoised images are generated using a second ANN as intermediate outputs of the second ANN during different training iterations. As such, the first ANN may learn to perform the denoising task in an unsupervised manner without requiring noise-free training images as the ground truth.

Abstract: Introduced here are diagnostic platforms able to verify subjects included in images in a time- and resource-efficient manner. By assessing the degree of consistency between multiple characterizations of an attribute that are derived from different data, a diagnostic platform can determine the likelihood that an image includes a given subject. An attribute could be characterized based on metadata that accompanies an image captured during a diagnostic session, information provided by an individual as part of the diagnostic session, an output produced by a classification model upon being applied to the image, or any combination thereof.

Abstract: In embodiments, a processing device generates a three-dimensional model of a dental site from scan data, the three-dimensional model comprising a representation of a tooth, wherein a portion of the three-dimensional model comprises an interfering surface that obscures a portion of the tooth. The processing device receives or generates an image of the tooth, wherein the image depicts the interfering surface. The processing device processes the image to generate a modified image, wherein the portion of the tooth that was obscured by the interfering surface in the image is shown in the modified image. The processing device updates the three-dimensional model of the dental site by replacing, using the modified image, the portion of the three-dimensional model that comprises the interfering surface that obscures the portion of the tooth, wherein the portion of the tooth that was obscured in the three-dimensional model is shown in an updated three-dimensional model.

Abstract: The disclosure relates to a method for diagnosing plant diseases and a plant disease diagnosis system. The method includes the following steps: acquiring a plant image; determining identification information according to the plant image, wherein the identification information includes at least one of species information and disease information; and extracting diagnostic information from a content management system according to the determined identification information, and outputting the diagnostic information.

Abstract: Apparatus and methods related to photography are provided. A computing device can receive an input image. An object detector of the computing device can determine an object region of interest of the input image that is associated with an object detected in the input image. A trained machine learning algorithm can determine an output photographic region of interest for the input image based on the object region of interest and the input image. The machine learning algorithm can be trained to identify an output photographic region of interest that is suitable for use by a photographic function for image generation. The computing device can generate an output related to the output photographic region of interest.

Abstract: Examples relate to a computer-implemented method, data processing apparatus and computer program for generating three-dimensional pose-estimation data. The method comprises obtaining video data of a plurality of cameras, the video data showing a movement of one or more persons, as perceived from a plurality of angles of observation. The method comprises generating two-dimensional pose-estimation data of the one or more persons using a machine-learning model that is suitable for performing two-dimensional pose-estimation based on the video data. The method comprises generating three-dimensional pose-estimation data of the one or more persons based on the two-dimensional pose-estimation data of the one or more persons. The two-dimensional and three-dimensional pose-estimation data is defined by one or more points on the body of the one or more persons.

Abstract: An image obtaining means including: a processor; a memory module communicatively coupled to the processor; a camera communicatively coupled to the processor; and the processor being configured to: receive, from the camera, image data; determine whether a predetermined object is present based on the image data; if a predetermined object is detected as present, determine if the predetermined object is moving or stationary; and if the predetermined object is stationary for at least a predetermined time period, record the duration that the predetermined object is stationary in the memory module.

Abstract: Disclosed are techniques for image processing that mitigate artifacts that otherwise appear when an IQVR camera moves with respect to a scene between image frames.