Abstract: This disclosure relates to detecting mirrored lidar returns and modifying lidar data using clustering techniques. In some examples, lidar data including a set of lidar points may be captured by a vehicle operating in an environment. The vehicle may combine the lidar data from numerous lidar devices to have a common frame of reference. In some examples, the vehicle may determine a cluster of the lidar points based on elevation data and azimuth data. The vehicle may compare range data of the clusters to determine whether such lidar points are indicative of a lidar return associated with an error (including multipath reflections). Based on determining that the difference in ranges of two lidar points is above a threshold value, the vehicle may determine how to use the data.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for reducing camera false detections. One of the methods includes providing, to a neural network of an image classifier that is trained to detect objects of two or more classification types, a feature vector for a respective training image; receiving, from the neural network, an output vector that indicates, for each of the two or more classification types, a likelihood that the respective training image depicts an object of the corresponding classification type; accessing, from two or more ground truth vectors each for one of the two or more classification types, a ground truth vector for the classification type of an object depicted in the training image; and adjusting one or more weights in the neural network using the output vector and the ground truth vector; and storing, in a memory, the image classifier.

Abstract: An inter-device positional relationship estimation apparatus includes an analysis information acquisition unit that acquires time-series analysis information, a recommendation processing unit that estimates unanalyzed detailed information in time-series analysis information applying a predetermined recommendation algorithm, an initial map creation unit that calculates a neighborhood determination point indicating a degree of correlation of detailed information between fixed devices to create an initial map, and a device relationship map creation unit that calculates neighborhood points indicating a degree of correlation of detailed information between respective devices including a moving device or/and a new fixed device to calculate a detection probability and a movement time between devices and create a device relationship map.

Abstract: In accordance with some embodiments, a portable electronic device includes: a display; one or more processors; memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or mor processors. The one or more programs include instructions for: detecting a user-defined gesture on the display while the portable electronic device is in a user-defined-gesture recording mode; creating a representation of the user-defined gesture; associating the representation of the user-defined gesture with an emoji character; and storing the representation of the user-defined gesture and the association of the representation of the user-defined gesture with the emoji character.

Abstract: A method for tracking and/or characterizing multiple objects in a sequence of images. The method includes: assigning a neural network to each object to be tracked; providing a memory that is shared by all neural networks; providing a local memory for each neural network, respectively; supplying images from the sequence, and/or details of these images, to each neural network; during the processing of each image and/or image detail by one of the neural networks, generating an address vector from at least one processing product of this neural network; based on this address vector, writing at least one further processing product of the neural network into the shared memory and/or into the local memory, and/or reading out data from this shared memory and/or local memory and further processing the data by the neural network.

Abstract: An image capture apparatus, remotely controllable by an operation apparatus, includes an image capturing unit capturing an image, a communication unit communicating with the operation apparatus, an image processing unit detecting an object in the image, and a control unit performing framing control of the image including the object, based on a framing operation from the operation apparatus, after the communication unit transmits the image to the operation apparatus, and a delay acquisition unit acquiring a delay time in the communication processing. The image processing unit calculates a framing control amount based on a predicted movement amount of the object in the delay, an actual movement amount of the object before and after the delay, and a framing operation amount from the operation apparatus. The control unit performs the framing control based on the framing control amount, and the delay acquisition unit acquires the delay time, before shooting, in advance.

Abstract: A multi-mode tracking method according to the present disclosure includes receiving, a sensor signal, obtaining a sensor-based location of a sports participant based on the sensor signal, obtaining a first credibility information related to a credibility of the sensor-based location, receiving a sports image captured at a camera disposed peripheral to a playfield, the sports image including the sports participant in the playfield, obtaining an image-based location of the sports participant, obtaining an second credibility information related to a credibility of the image-based location, wherein the credibility of the image-based location is related to an occlusion related to the sports participant, calculating an weight value based on the first credibility information and the second credibility information, calculating a location of the sports participant based on the weight value.

Abstract: A method and a device for image processing, an imaging system, and a storage medium are provided. The method includes: obtaining an initial video shot by a photographing apparatus at a fixed angle of view, where the initial video includes a plurality of frames of images; cropping, based on a preset trajectory, the plurality of frames of images in the initial video; and combining the cropped plurality of frames of images to obtain a target video. In this way, a video with a smoothly moving field of view may be obtained based on a video shot at a fixed angle of view, thereby improving a user's shooting experience.

Abstract: Systems and methods for performing semantic segmentation of three-dimensional data are provided. In one example embodiment, a computing system can be configured to obtain sensor data including three-dimensional data associated with an environment. The three-dimensional data can include a plurality of points and can be associated with one or more times. The computing system can be configured to determine data indicative of a two-dimensional voxel representation associated with the environment based at least in part on the three-dimensional data. The computing system can be configured to determine a classification for each point of the plurality of points within the three-dimensional data based at least in part on the two-dimensional voxel representation associated with the environment and a machine-learned semantic segmentation model. The computing system can be configured to initiate one or more actions based at least in part on the per-point classifications.

Abstract: A method includes obtaining a first imaging scan and a second imaging scan of a single subject brain. The first imaging scan is converted to a first dataset, and the second imaging scan is converted to a second dataset. A sequence-adaptive multimodal segmentation algorithm is applied to the first dataset and the second dataset. The sequence-adaptive multimodal segmentation algorithm performs automatic intensity-based tissue classification to generate a first labelled dataset and a second labeled dataset. The first labeled dataset and the second labeled dataset are automatically co-registered to each other to generate a transformation matrix based on the first labeled dataset and the second labeled dataset. The transformation matrix is applied to align the first dataset and the second dataset.

Abstract: A method for processing image data according to an aspect of the present disclosure includes obtaining first image data indicating a hyperspectral image of a target captured in first background light, generating, on a basis of the first image data, first spectral data indicating an estimated spectrum of the first background light, and generating, from the first image data, at least one piece of second image data indicating at least one image of the target in at least one type of second background light, which is different from the first background light, using at least one piece of second spectral data indicating at least one spectrum of the at least one type of second background light and the first spectral data.

Abstract: Disclosed herein are an apparatus for estimating a camera pose using multi-view images of a 2D array structure and a method using the same. The method performed by the apparatus includes acquiring multi-view images from a 2D array camera system, forming a 2D image link structure corresponding to the multi-view images in consideration of the geometric structure of the camera system, estimating an initial camera pose based on an adjacent image extracted from the 2D image link structure and a pair of corresponding feature points, and estimating a final camera pose by reconstructing a 3D structure based on the initial camera pose and performing correction so as to minimize a reprojection error of the reconstructed 3D structure.

Abstract: An apparatus and method for automatic license plate recognition of a vehicle, the apparatus includes a user device configured to receive a user input containing reference indicators and associate a right to the reference indicators, an image capturing device configured to capture vehicle images of a vehicle containing a license plate region, an image processing module configured to receive the vehicle images from the image capturing device and process the vehicle images, a computer vision module configured to determine an obscurity of the license plate region, and identify indicators as a function of the license plate region, an image inpainting module configured to reconstruct the license plate region based on the obscurity of the license plate using an image inpainting technique, and a validation module configured to validate the indicators as a function of the reference indicators and grant the user the right as a function of the validation.

Abstract: An imaging support method includes acquiring a focal length of an imaging apparatus, detecting a subject image position of a target subject image showing a target subject in a captured image obtained by capturing an imaging region including the target subject by the imaging apparatus, and performing registration control for setting the detected subject image position to a specific position in the captured image by increasing a ratio of application of an image region movement component compared to a lens moving mechanism in a case where the acquired focal length is greater than or equal to a threshold value, and increasing a ratio of application of the lens moving mechanism compared to the image region movement component in a case where the focal length is less than the threshold value.

Abstract: The disclosure provides an LED light source recognition method, device, apparatus, and medium based on deep learning, belonging to the field of indoor positioning and navigation technology. The method includes the following. In an LED lighting environment, a spline frame is obtained through a CMOS camera, in which the spline frame is an image having dark stripes. The spline frame is input to a target detection model, a dark stripe detection result output by the target detection model is obtained. The multiple rectangular boxes are preprocessed, and based on the predicted classification corresponding to each preprocessed rectangular box, an image feature encoding sequence of the spline frame is determined. The image feature encoding sequence is compared with a light source feature encoding sequence corresponding to each LED light source to determine the light source feature encoding sequence that best matches the spline frame.

Abstract: Systems and methods for recommending a media asset relating to a character unknown to a user are provided herein. The systems and methods may receive a first media asset viewed by a user, determine a first character in the first media asset, and include the first character in the character viewing profile for the user. Further, the systems and methods may receive a second media asset, including the first character, determine that a second character appears in the second media asset, and determine that the second character is not in the character viewing profile. The systems and methods may recommend a third media asset to the user for the second character.

Abstract: Disclosed herein is a method for operating an audio device, in particular for carrying out a morphing process, wherein audio output data are generated from audio input data containing at least one sample and/or at least one block of samples by means of at least one function stored in a data device, wherein at least one parameter of the at least one function is changeable and/or is changed in operation of the audio device, wherein at least one item of topicality information of the at least one parameter, which information indicates a state of change of the at least one parameter, is updated after the change has been made, wherein the audio output data is generated depending on the topicality information of the at least one parameter by means of the at least one changed or the at least one unchanged parameter of the at least one function.

Abstract: A method comprising: receiving, as input, training images, wherein at least a majority of the training images represent normal data instances; receiving, as input, a target image; extracting (i) a set of feature representations from a plurality of image locations within each of the training images, and (ii) target feature representations from a plurality of target image locations within the target image; calculating, with respect to a target image location of the plurality of target image locations in the target image, a distance between (iii) the target feature representation of the target image location, and (iv) a subset from the set of feature representations comprising the k nearest the feature representations to the target feature representation; and determining that the target image location is anomalous, when the calculated distance exceeds a predetermined threshold.

Abstract: A system and method are presented that scans a vector graphics file for image elements that appear over a background image. The image elements represent individual objects and movement or actions for those objects. The image elements, each possibly represented by multiple shapes, arrows, lines, or shading, are identified and parsed into a standard class definition. In one embodiment, the class definitions and the rule set applied to perform this parsing are pre-defined according to the content of the SVG file. The image elements are then redrawn over a new background image, which has multiple segments that correspond to the original background image but where each segment has a different segment transformation that defines translation and/or scaling required to move between the image segments. The image elements are then scaled and translated according to the multiple segment transformations, and then are presented over the new background image.

Abstract: A machine learning scheme can be trained on a set of labeled training images of a subject in different poses, with different textures, and with different background environments. The label or marker data of the subject may be stored as metadata to a 3D model of the subject or rendered images of the subject. The machine learning scheme may be implemented as a supervised learning scheme that can automatically identify the labeled data to create a classification model. The classification model can classify a depicted subject in many different environments and arrangements (e.g., poses).

Abstract: Systems, methods, and computer-readable media are provided for wearable navigation systems. In some examples, a wearable navigation system may determine the distance between an object and one or more distance sensors based on signals received from an environment, where the distance sensors are part of a sensor package. In some aspects, the wearable navigation system may determine, using an inertial measurement unit, a position, speed, and acceleration of the wearable navigation system, where the inertial measurement unit is part of the sensor package. In some cases, the wearable navigation system may determine, by a Central Processing Unit (CPU) coupled to the sensor package, information associated with the object with respect to the wearable navigation system. In some instances, the wearable navigation system may generate, by an actuator system coupled to the sensor package, feedback signals in response to information associated with the object.

Abstract: A method for analyzing a sensor assembly configuration of an at least semi-autonomous robot includes determining a plurality of spatial segments that spatially subdivide an environment of the robot. The method further includes determining an available individual component performance of a respective individual component of the sensor assembly configuration in relation to the spatial segments. The method further includes determining a sensor assembly requirement that must be satisfied by the sensor assembly configuration in relation to the spatial segments. A linear optimization function with parameters that include the spatial segments, the individual component performances, and the sensor assembly requirements is generated and then solved according to the method. The solution of the linear optimization function indicates if the environment of the robot is configured to be captured by the sensor assembly configuration in accordance with the sensor assembly requirement.

Abstract: Systems and methods to determine an activity associated with an object of interest are presented. The method includes receiving a sequence of a plurality of frames that capture an area of interest during a threshold period of time, wherein an object of interest is present in the area of interest during the threshold period of time. The method also includes analyzing the plurality of frames to detect the object of interest. The method also includes tracking the object of interest in the area of interest during the threshold period of time. The method also includes extracting data indicative of the object of interest. The method also includes predicting, based on the data indicative of the object of interest, an activity associated with the object of interest.

Abstract: A device detects anomalies on an input stream of events. Each event includes a value. The device has: a modeling unit configured to represent each event into at least one histogram over a predefined tree data structure to obtain a model of the input stream; a change detection and adaptation module configured to detect a long term change of the distribution of the events in the input stream based on the model of the input stream; and a scoring unit configured to calculate an anomaly score representing the probability of an anomaly in the input stream of events based on the model of the input stream and the detected long term change.

Abstract: A device and method with object recognition is included. In one general aspect, an electronic device includes a camera sensor configured to capture a first image of a scene, the camera sensor is configured to perform at least one type of physical camera motion relative to the electronic device, the at least one type of physical camera motion includes rolling, panning, tilting, or zooming the camera sensor relative to the electronic device, and a processor configured to control the camera sensor to perform a physical motion of the physical camera motion type based on detecting an object in the first image, acquire a second image captured using the camera sensor as adjusted based on the performed physical motion, and recognize the object in the second image.

Abstract: In some embodiments, a method can include augmenting a set of images of collectables to generate a set of synthetic images of collectables. The method can further include combining the set of images of collectables and the set of synthetic images of collectables to produce a training set. The method can further include training a set of machine learning models based on the training set. Each machine learning model from the set of machine learning models can generate a grade for an image attribute from a set of image attributes. The set of image attributes can include an edge, a corner, a center, or a surface. The method can further include executing, after training, the set of machine learning models to generate a set of grades for an image of collectable not included in the training set.

Abstract: A surveying instrument for executing a relocation functionality, which determines first coordinates of a stationary target point associated with the start signal, in response to a start signal, a first actuator and a second actuator are controlled such that the stationary target point remains within a detection area of a tracking unit of the surveying instrument, determines second coordinates of the stationary target point, receives an end signal, wherein the second coordinates of the stationary target point are associated with the end signal, and based at least in part on the first and second coordinates of the stationary target point, and determines a relative pose of the surveying instrument with respect to a first setup location and a second setup location, wherein the first setup location is associated with the first coordinates and the second setup location is associated with the second coordinate.

Abstract: A method for reducing camera motion blur comprises, before acquiring an image frame for a video stream, a camera measurement unit measuring data related to a camera module motion during a time window; determining camera module motion based on the measured data and predicting a camera motion blur during acquisition of the image frame based at least on the determined camera module motion and the lens projection model; determining whether the predicted camera motion blur exceeds a threshold; in response to determining that the predicted camera motion blur exceeds the threshold, determining a reduction of the provisional exposure time determined to acquire the image frame so that the predicted camera motion blur reaches the threshold, determining whether a corresponding increase in the provisional gain determined to acquire the image frame is below a maximum gain value, adjusting the provisional exposure time and gain, and acquiring the image frame.

Abstract: An estimation apparatus includes one or more hardware processors configured to detect a first object included in time-series images, and generate a tracking trajectory of the first object; detect a state change event indicating an appearance, a disappearance, a bend, or a stay of the tracking trajectory, and extract a coordinate of the first object in which the state change event has occurred; and estimate a determination region based on the coordinate. Accordingly, the estimation apparatus is able to estimate more accurately an image region to be analyzed.

Abstract: Various implementations disclosed herein include devices, systems, and methods for obfuscating location data associated with a physical environment. In some implementations, a method includes obtaining, via an environmental sensor, environmental data corresponding to a physical environment. A first portion of the environmental data that corresponds to a first location is identified. In response to the first location being of a first location type, location data indicative of the first location is obfuscated from the environmental data by modifying the first portion of the environmental data.

Abstract: A process to partition a video feed to segment live player activity includes receiving, on a first recurring basis, a transmission of a central video feed from a first camera. The central video feed is calibrated against a spatial region represented in at least two dimensions that is encompassed by the central video feed. The process includes receiving, on a second recurring basis, a respective time-stamped position information from each tracking device in a plurality of tracking devices. Each tracking device is worn by a corresponding subject on the spatial region and transmits positional information that describes a time-stamped position of the corresponding subject in the spatial region. The process uses the received information and the calibration to define a first sub-view of the central video feed associated with a first subject. The first sub-view comprises a corresponding sub-frame associated with the first subject.

Abstract: The present disclosure discloses an electro-hydraulic varifocal lens-based method for tracking a 3D trajectory of a moving object. The method includes the following steps of: (1) obtaining a functional relation between a focusing control current and camera's intrinsic parameters; (2) obtaining a functional relation between focusing control currents of the electro-hydraulic varifocal lens and an optimal object distance; (3) initializing an object tracking algorithm, and taking an object tracking box as a subsequent focusing window; (4) carrying out first autofocusing, recording a focusing control current value after the autofocusing is completed, as well as a size and center point coordinates of the object tracking box; (5) calculating and recording coordinates of the object in 3D space; and (6) repeating steps (4) and (5) for the same object, and sequentially connecting the recorded coordinates of the object in 3D space into a trajectory.

Abstract: Real-time light-detection and ranging point decimation is provided. A system of a vehicle can allocate, based on a resolution for data points, a region in the memory for a tree data structure. The region can have a fixed-size for a duration of a flight performed by the vehicle. The system can receive, subsequent to allocating the region in the memory, first points captured by a sensor of the vehicle at a first resolution, and convert the points to second points having a second resolution that is less than the first resolution. The system can then provide, for storage in the region in the memory, the tree data structure that stores the second points. The tree data structure can be used in a terrain map generation function.

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: The present invention relates to system and method that detects objects and assigns trackers to each object and maps it to detections to its corresponding Geo-Coordinates. Based on the number of detections, it assigns new trackers and counts the objects identified using optimal and efficient computation. The present system requires just one detection and its location to create a tracker accurately. Also, if an object is missing in any given location, the present system identifies the Geo-Coordinates and shows the image along with its details.

Abstract: Aspects of the present disclosure relate generally to a vision system that detects persons exiting an environment, comprising detecting, by a processor using a plurality of image frames from at least one sensor, persons that exited the environment during a first period of time. The vision system further determines, by the processor, an exit count for the first period of time and retrieves, from a database, historical egress data comprising a detected historic exit count and a corrected historic exit count of the environment for a second period of time corresponding to the first period of time. The vision system calculates an error rate for the second period of time based on a ratio of the detected historic exit count and the corrected historic exit count, determines and stores a corrected exit count for the first period of time by adjusting the exit count using the error rate.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for preserving privacy in surveillance. The methods, systems, and apparatus include actions of determining a state of a monitoring system, determining an exclusion zone that is shown in a video, determining whether to obfuscate at least a portion of the video based on the exclusion zone and the state of the monitoring system, and obfuscating at least the portion of the video.

Abstract: A method and system for setting a vocal collision queue is disclosed. A plurality of incoming audio streams associated with a plurality of user devices currently involved in a communication session is recorded. There is an overlap between portions of two or more of the audio streams in the communication session that is identified. A queue for the two or more identified audio streams is determined. The queue includes a sequence to the identified audio streams. Recordings corresponding to the overlapping portions of the identified audio stream are retrieved. The retrieved recordings are provided to the user devices in the communication session. The retrieved recordings are played sequentially in accordance with the determined sequence.

Abstract: A hyperspectral imaging device (100) is provided comprising an input (102) for receiving a light field from a scene (106), an encoder (108), at least one dispersive element (110, 112), at least one array detector (114, 110) and a processor (118). The encoder (108) is arranged to receive at least a portion of the light field from the input (102) and transform it to provide a first and second encoded light (120, 122) field having different spatial patterns. At least one dispersive element (110, 112) is arranged to apply spectral shear to the first and second encoded light fields (120, 122) respectively to provide first and second sheared light fields (124, 126). At least one array detector (114, 116) is arranged to detect the first and second sheared light fields (124, 126). The processor (118) is arranged to process an output from the at least one array detector (114, 116) to determine a datacube (128) corresponding to a hyperspectral image of the scene.

Abstract: Systems and methods for a vehicle are provided. In one embodiment, a method includes: receiving image data defining a plurality of images associated with an environment of the vehicle; determining, by a processor, feature points within at least one image of the plurality of images; selecting, by the processor, a subset of the feature points as ground points based on a fixed two dimensional image road mask and a three dimensional region; determining, by the processor, a ground plane based on the subset of feature points; determining, by the processor, a ground normal vector from the ground plane; determining, by the processor, a camera to ground alignment value based on the ground normal vector; and generating, by the processor, second image data based on the camera to ground alignment value.

Abstract: One embodiment is directed to a system for enabling two or more users to interact within a virtual world comprising virtual world data, comprising a computer network comprising one or more computing devices, the one or more computing devices comprising memory, processing circuitry, and software stored at least in part in the memory and executable by the processing circuitry to process at least a portion of the virtual world data; wherein at least a first portion of the virtual world data originates from a first user virtual world local to a first user, and wherein the computer network is operable to transmit the first portion to a user device for presentation to a second user, such that the second user may experience the first portion from the location of the second user, such that aspects of the first user virtual world are effectively passed to the second user.

Abstract: The present disclosure relates to a broadcast directing method, apparatus and system, and a non-transitory computer-storable medium, which relate to the technical field of computers. The broadcast directing method includes acquiring a reference video stream from a reference camera position; performing event recognition on the reference video stream to obtain at least one reference event frame interval; determining a local sequence of each reference event frame interval according to a correspondence relationship between the event and a camera position identification, the local sequence comprising a camera position identification of each frame image of a video to be played corresponding to the reference event frame interval and a frame identification corresponding to the camera position identification; generating a broadcast directing sequence according to the local sequence; and generating a broadcast directing video according to the broadcast directing sequence and a video stream of a camera position.

Abstract: Devices, systems, and methods for determining whether a container is empty in the context of robotic picking solutions. The system includes a plurality of sensors configured to gather container data regarding a container at a first location, wherein the container data includes at least two of weight data related to the container, depth data related to the container, and color sensor data related to the container, and a processor configured to execute instructions stored on a memory to provide a sensor fusion module configured to process the received container data to determine whether the container is empty.

Abstract: This document describes methods and systems directed at history-based identification of incompatible tracks. The historical trajectory of tracks can be advantageous to accurately determine whether tracks originating from different sensors identify the same object or different objects. However, recording historical data of several tracks may consume vast amounts of memory or computing resources, and related computations may become complex. The methods and systems described herein enable a sensor fusion system of an automobile or other vehicle to consider historical data when associating and pairing tracks, without requiring large amounts of memory and without tying up other computing resources.

Abstract: The invention relates to a detector device having pixels in a two-dimensional arrangement, the detector device comprising row pixels, a row readout wiring for each row of said row pixels and a column select wiring for each column of said row pixels, as well as column pixels, a column readout wiring for each column of said column pixels and a row select wiring for each row of said column pixels. Analog-to-digital converters adapted for row-parallel and column-parallel operation are connected to the readout wirings, and simultaneously one or more columns of row pixels for row parallel readout and one or more rows of column pixels for column parallel readout can be selected. The invention further relates to a positioning code and to a position detecting method.

Abstract: A device for orientation and position detection of markings in three-dimensional space includes a marking arrangement, an optical image capture unit and an evaluation unit. The marking arrangement comprises two or more marking units each having three or more illuminating means arranged along a path. At least one marking unit belongs to a first marking unit type having at least three markings, and at least one other marking unit belongs to a second marking unit type having exactly two markings and a communication element arranged therebetween. At least one marking unit of the first type and at least one marking unit of the second type are arranged in a non-coplanar manner. The evaluation unit is adapted to recognize the marking units of the different types and to unambiguously determine the orientation and the position of the marking arrangement using the markings of recognized marking units of the different types.

Abstract: A device configured to transcribe an appearance of a human being, including a common housing holding an image capturing sensor, a computing device having a data processor, and a computer program product including a first machine learning model trained for detecting and labeling human beings, a second machine learning model trained for detecting appearances of human beings and a transcription module to transcribe the detected appearances of human beings to text.

Abstract: Disclosed is a washing machine. The washing machine includes an image capturer, a storage, and a controller. When the washing machine is in a standby state, if photographed data is identified to be registered user data, the controller recognizes a photographed user as a registered user and performs control such that a door is opened. If the photographed data is identified to be child data, the controller performs control such that, when the door is in an open state, the door is not closed and such that, when the door is in a closed state, the door is not opened.

Abstract: An AI-based pre-training model determination system is proposed. When determination type information is input, the AI-based pre-training model determination system extracts a candidate model among a plurality of learning models on the basis of determination type information, and the candidate model determines new training data. An uppermost candidate model whose determination accuracy is greater than or equal to a first reference value preset on the basis of the determination accuracy of the candidate model is determined as a pre-training model for generation of a new learning model, thereby improving the determination accuracy of the new learning model.

Abstract: A system and a method for processing an image include inputting the image to a neural network configured to: obtain a plurality of feature maps, each feature map having a respective resolution and a respective depth, perform a classification on each feature map to deliver, for each feature map: the type of at least one object visible on the image, the position and shape in the image of at least one two-dimensional bounding box surrounding the at least one object, a plurality of rotation hypotheses for the at least one object.