Abstract: A method for stabilizing an image based on artificial intelligence includes acquiring tremor detection data with respect to the image, the tremor detection data acquired from two or more sensors; outputting stabilization data for compensating for an image shaking, the stabilization data outputted using an artificial neural network (ANN) model trained to output the stabilization data based on the tremor detection data; and compensating for the image shaking using the stabilization data. A camera module includes a lens; an image sensor to output an image captured through the lens; two or more sensors to output tremor detection data with respect to the image; a controller to output stabilization data based on the tremor detection data using an ANN model; and a stabilization unit to compensate for an image shaking using the stabilization data. The ANN model is trained to output the stabilization data based on the tremor detection data.

Abstract: Embodiments provide a lens moving apparatus including a bobbin in which a lens is mounted, a first coil and a magnet configured to electromagnetically interact with each other so as to move the bobbin, a housing configured to accommodate the bobbin therein, an elastic member including an inner frame coupled to the bobbin, an outer frame coupled to the housing, and a frame connection portion configured to connect the inner frame and the outer frame to each other, and a support member connected to the elastic member and configured to support the housing, and the outer frame includes a first coupling portion coupled to the housing, a second coupling portion coupled to the support member, the second coupling portion being spaced apart from the first coupling portion, and a single connection portion configured to connect the first coupling portion and the second coupling portion to each other.

Abstract: Provided is a neural processing unit that performs application-work including a first neural network operation, the neural processing unit includes a first processing core configured to execute the first neural network operation, a hardware block reconfigurable as a hardware core configured to perform hardware block-work, and at least one processor configured to execute computer-readable instructions to distribute a part of the application-work as the hardware block-work to the hardware block based on a first workload of the first processing core.

Abstract: A method for generating a human-computer interactive abstract image includes: S1: obtaining and preprocessing the original abstract images used as a training dataset B to obtain edge shape feature maps used as a training dataset A; S2: using the training dataset A and the training dataset B as cycle generative objects of a Cycle-GAN model, and training the Cycle-GAN model to capture a mapping relationship between the edge shape feature maps and the original abstract images; S3: obtaining a line shape image drawn by a user; and S4: according to the mapping relationship, intercepting a generative part in the Cycle-GAN model that the dataset B is generated from the dataset A, discarding a cycle generative part and a discrimination part in the Cycle-GAN model, and generating a complete abstract image based on the line shape image to generate the human-computer interactive abstract image.

Abstract: A vision sensor includes a pixel array comprising pixels arranged in a matrix, an event detection circuit, an event rate controller, and an interface circuit. Each pixel is configured to generate an electrical signal in response to detecting a change in incident light intensity. The event detection circuit detects whether a change in incident light intensity has occurred at any pixels, based on processing electrical signals received from one or more pixels, and generates one or more event signals corresponding to one or more pixels at which a change in intensity of incident light is determined to have occurred. The event rate controller selects a selection of one or more event signals corresponding to a region of interest on the pixel array as one or more output event signals. The interface circuit communicates with an external processor to transmit the one or more output event signals to the external processor.

Abstract: A neural network recognition method includes obtaining a first neural network that includes layers and a second neural network that includes a layer connected to the first neural network, actuating a processor to compute a first feature map from input data based on a layer of the first neural network, compute a second feature map from the input data based on the layer connected to the first neural network in the second neural network, and generate a recognition result based on the first neural network from an intermediate feature map computed by applying an element-wise operation to the first feature map and the second feature map.

Abstract: A focus detection device includes: an imaging unit having a first pixel and a second pixel each of which receives light transmitted through an optical system and outputs signal used for focus detection; an input unit to which first information regarding a position on an image plane and an exit pupil distance of the optical system is input; a selection unit that selects a first focus detection based on the signal having been output from the first pixel or a second focus detection based on the signal having been output from the second pixel, based on the first information having been input to the input unit; and a focus detection unit that performs the first focus detection or the second focus detection based on a selection by the selection unit.

Abstract: A computing system retrieves ball-by-ball data for a plurality of sporting events. The computing system generates a trained neural network based on ball-by-ball data supplemented with ball-by-ball data with ball-by-ball match context features and personalized embeddings based on a batsman and a bowler for each delivery. The computing system receives a target batsman and a target bowler for a pitch to be delivered in a target event. The computing system identifies target ball-by-ball data for a window of pitches preceding the to be delivered pitch. The computing system retrieves historical ball-by-ball data for each of the target batsman and the target bowler. The computing system generates personalized embeddings for both the target batsman and the target bowler based on the historical ball-by-ball data. The computing system predicts a shot type for the pitch to be delivered based on the target ball-by-ball data and the personalized embeddings.

Abstract: Apparatus, systems, and methods to deliver point of care alerts for radiological findings are disclosed. An example imaging apparatus includes an image data store, an image quality checker, and a training learning network. The example image data store is to store image data acquired using the imaging apparatus. The example image quality checker is to evaluate image data from the image data store in comparison to an image quality measure. The example trained learning network is to process the image data to identify a clinical finding in the image data, the identification of a clinical finding to trigger a notification at the imaging apparatus to notify a healthcare practitioner regarding the clinical finding and prompt a responsive action with respect to a patient associated with the image data.

Abstract: A convolutional neural network includes a pooling unit. The pooling unit performs pooling operations between convolution layers of the convolutional neural network. The pooling unit includes hardware blocks that promote computational and area efficiency in the convolutional neural network.

Abstract: Described are system, method, and computer-program product embodiments for developing an object detection model. The object detection model may detect a physical object in an image of a real world environment. A system can automatically generate a plurality of synthetic images. The synthetic images can be generated by randomly selecting parameters of the environmental features, camera intrinsics, and a target object. The system may automatically annotate the synthetic images to identify the target object. In some embodiments, the annotations can include information about the target object determined at the time the synthetic images are generated. The object detection model can be trained to detect the physical object using the annotated synthetic images. The trained object detection model can be validated and tested using at least one image of a real world environment. The image(s) of the real world environment may or may not include the physical object.

Abstract: Disclosed is a method of tracking multiple objects included in an image using a multi-object tracking apparatus including an integrated similarity neural network, the method including setting a tracking area in an input image, extracting at least one object candidate for a target object from the tracking area; extracting reference features for the target object, the object candidate, and the tracking area, selecting two of the target object, the object candidate, and the tracking area to evaluate similarity based on the reference features; allocating the object candidate to the target object on the basis of the evaluated similarity; and tracking the target object on the basis of a location of the allocated object candidate.

Abstract: A system implements a key value memory network including a key matrix with key vectors learned from training static feature data and time-series feature data, a value matrix with value vectors representing time-series trends, and an input layer to receive, for a target entity, input data comprising a concatenation of static feature data of the target entity, time-specific feature data, and time-series feature data for the target entity. The key value memory network also includes an entity-embedding layer to generate an input vector from the input data, a key-addressing layer to generate a weight vector indicating similarities between the key vectors and the input vector, a value-reading layer to compute a context vector from the weight and value vectors, and an output layer to generate predicted time-series data for a target metric of the target entity by applying a continuous activation function to the context vector and the input vector.

Abstract: An image processing apparatus which superimposes an indicator image on a picked-up image, with inhibition of a subject from deteriorating in visibility. At least one processor of the image processing apparatus executes the set of instructions to: specify two points of a subject in a picked-up image; acquire three-dimensional positional information about the two points; detect a change about a state of the image processing apparatus; generate an indicator image corresponding to a length between the two points and the change, based on the three-dimensional positional information and the change; and superimpose the indicator image onto the picked-up image, to acquire a retouched image.

Abstract: Provided by the present invention is a picture focusing method, comprising: acquiring a video picture, and, according to an instruction command, determining a target object in the video picture; determining picture depth information of a video picture region in which the target object is located; acquiring parameter information of a camera corresponding to multiple video picture frames; using the picture depth information of the video picture region and the parameter information of the camera as the input of a preset focusing neural network model; and using focus adjustment information of the camera to perform a focus adjustment operation on the camera.

Abstract: Provided are methods for object tracking, which can include receiving sensor data characterizing respective detected objects. The methods can also include generating a data structure based on the data characterizing the respective detected objects. The data structure can include a graph of nodes representing states of the objects and edges representing hypothetical transitions in states of the objects. The methods can also include applying a predictive model to the data structure. The predictive model can be trained to receive the state as inputs and produce an identification of a set of nodes and edges corresponding to the one of the respective detected objects. The methods can further include providing data based on the identification of the set of nodes and edges to a planning system of the vehicle and causing the vehicle to operate based on providing the data. Systems and computer program products are also provided.

Abstract: A solid-state imaging device including an imaging area where a plurality of unit pixels are disposed to capture a color image, wherein each of the unit pixels includes: a plurality of photoelectric conversion portions; a plurality of transfer gates, each of which is disposed in each of the photoelectric conversion portions to transfer signal charges from the photoelectric conversion portion; and a floating diffusion to which the signal charges are transferred from the plurality of the photoelectric conversion portions by the plurality of the transfer gates, wherein the plurality of the photoelectric conversion portions receive light of the same color to generate the signal charges, and wherein the signal charges transferred from the plurality of the photoelectric conversion portions to the floating diffusion are added to be output as an electrical signal.

Abstract: Techniques are provided for increasing the accuracy of automated classifications produced by a machine learning engine. Specifically, the classification produced by a machine learning engine for one photo-realistic image is adjusted based on the classifications produced by the machine learning engine for other photo-realistic images that correspond to the same portion of a 3D model that has been generated based on the photo-realistic images. Techniques are also provided for using the classifications of the photo-realistic images that were used to create a 3D model to automatically classify portions of the 3D model. The classifications assigned to the various portions of the 3D model in this manner may also be used as a factor for automatically segmenting the 3D model.

Abstract: A field of view of a captured image of a lensless camera can be controlled, and a configuration of generating a restored image including part of an imaging region is realized. Included is a signal processing unit that receives observed image signals as output of an image sensor of a lensless camera to generate a restored image of a restored image region including part of a captured image region of the lensless camera.

Abstract: An apparatus for classifying an image according to an embodiment includes a fake image generation module receiving a classification target image and a fake image in a form in which only a background exists and no specific object exists in the classification target image, a difference of images vector generation module generating a difference of images between the classification target image and the fake image and a difference of images vector by converting the generated difference of images into preset one-dimensional matrix data, a difference of feature vectors generation module generating a difference of feature vectors between a feature vector generated based on the classification target image and a feature vector generated based on the fake image, and an image classification module classifying the classification target image based on the difference of images vector, the feature vector generated based on the classification target image, and the difference of feature vectors.