Abstract: A matrix multiplier includes an input vector scaler configured to generate a first scaled input vector based on a first input vector and a plurality of quantization scale coefficients, a first data type converter configured to generate a first fixed-point scaled input vector based on the first scaled input vector, a processing element array including a first processing element configured to generate a first fixed-point output element based on the first fixed-point scaled input vector and first plurality of quantization sign values and a second processing element configured to generate a second fixed-point output element based on the first fixed-point scaled input vector and second plurality of quantization sign values, and a second data type converter configured to generate first and second output elements by converting data type of the first and second fixed-point output elements, and to output a first output vector including the first and second output elements.

Abstract: A matrix multiplier includes an input vector scaler generating a first quantization scaled input vector based on a first input vector, a plurality of common scale coefficients, and first-to-Rth multiplication scale coefficients, a first data type converter generating a first fixed point quantization scaled input vector based on the first quantization scaled input vector, an element array comprising a first processing element generating a first fixed point output element based on the first fixed point quantization scaled input vector and first plurality of quantization sign bits, and a second processing element generating a second fixed point output element based on the first fixed point quantization scaled input vector and second plurality of quantization sign bits, and a second data type converter generating and outputting first and second output elements by converting data types of the first and second fixed point output elements.

Abstract: A method for playing streaming content includes requesting a playlist from a streaming server for providing a streaming service; receiving, in response to the request, a streaming fragment template provided by the streaming server; separating the playlist, a manifest file, and track information included in the streaming fragment template; and processing the play of the streaming content requested from a player module by using at least one of the playlist, the manifest file, and the track information.

Abstract: An annotation assistant for semi-automatic creation of annotated machine learning training data comprises a graphical annotation interface provider that provides a graphical annotation interface for displaying a sequence of tokens to be annotated and allowing a span selection by a user. An automatic span tagger comprises an artificial neural network configured to assign class probabilities to the tokens of the sequence of tokens and a processor-based path-scoring algorithm for computing scores for semantic attribute paths of the span selection. The graphical annotation interface is further configured for displaying the scored semantic attribute paths, and for receiving a selection among the scored semantic attribute paths from the user.

Abstract: Disclosed is an accelerator device which includes an interface circuit that communicates with an external device, a memory that stores first data received through the interface circuit, a polar encoder that performs polar encoding with respect to the first data provided from the memory and to output a result of the polar encoding as second data, and an accelerator core that loads the second data. The first data are compressed weight data, the second data are decompressed weight data, the accelerator core is configured to perform machine learning-based inference based on the second data, and the first data are variable in length.

Abstract: At least one embodiment provides a computing device including: a controller that receives first input data of a first data type and second input data of a second data type different from the first data type, and outputs a first signal representing the first data type, a second signal representing the second data type, and a clock signal based on the number of bits of the first input data and the second input data, and a computing circuit that performs a multiplication computation the first input data and the second input data based on the first signal, the second signal, and the clock signal and generates output data.

Abstract: A map generation method including receiving a map editing request for a specific floor among a plurality of floors of a building, providing an editing interface on a display unit of an electronic device in response to the map editing request, the editing interface including at least a part of a specific map corresponding to the specific floor, specifying at least one node group allocatable on the specific map based on first node rules, the first node rules corresponding to spatial characteristics of the specific floor, and performing a node placement process such that first nodes included among the at least one node group are placed on the specific map.

Abstract: A map generation method and system for robot operation. The map generation method including receiving a map editing request for a specific floor among a plurality of floors of a building, providing an editing interface on a display unit of an electronic device in response to the map editing request, the editing interface including at least a part of a specific map corresponding to the specific floor, allocating at least one graphic object on the specific map based on editing information received from the electronic device, and updating the specific map on a cloud server based on completion of the allocating such that robots travel through the specific floor according to an attribute of the at least one graphic object.

Abstract: A training system includes: a transformer module having the transformer architecture and configured to perform a vision task; and a training module configured to: receive a training image having a predetermined resolution; determine N windows of tokens of pixels in the training image and mask the tokens of all of the other pixels of the training image that are outside of the N windows, where N is an integer greater than or equal to 2; input the N windows of tokens to the transformer module; train the transformer module based on an output of the transformer module generated based on the N windows of tokens; and test the transformer module using a test image having the predetermined resolution.

Abstract: A method for preprocessing a set of feasible transfers within a multimodal transportation network of predetermined stations, comprising, for each trip in the multimodal transportation network hereafter called origin trip: (a) for each station (pti) of the origin trip (t), computing at this station (pti) an earliest arrival/change time associated with all transportation modes (m) of the multimodal transportation network; (b) for at least one transfer of the set of feasible transfers from a station (pti) on the origin trip (t) to a reachable station (puj) on a target trip (u), computing, at each station (puk>j) of the target trip (u) after the reachable station (puj), a value of the earliest arrival/change time specifically associated with the transportation mode (mu) of the multimodal transportation network used by the target trip (u); (c) removing the transfer only if determining that each computed value of the earliest arrival/change time is not improved by the transfer; (d) outputting the set of feasible

Abstract: Disclosed is an accelerator performing an accumulation operation on a plurality of data, each being a floating point type. A method of operating the accelerator includes loading first data, finding a first exponent, which is a maximum value among exponents of the first data, generating aligned first fractions by performing a bit shift on first fractions of the first data based on the first exponent, and generating a first accumulated value by an accumulation operation on the aligned first fractions, loading second data, finding a second exponent, which is a maximum value among exponents of the second data, and generating a first aligned accumulated value by a bit shift on the first accumulated value, generating aligned second fractions by a bit shift on second fractions of the second data, and generating a second accumulated value by an accumulation operation on the aligned second fractions and the first aligned accumulated value.

Abstract: Provided is a method for providing augmented content, wherein whether a user terminal is located in a preset unit space is determined on the basis of the location of the user terminal, and if the user terminal is determined to be located in the unit space, an image captured by a camera is displayed through an augmented reality (AR) view by being augmented with content associated with the unit space.

Abstract: A position recognition method and a system based on visual information processing are disclosed A position recognition method according to one embodiment including the steps of: generating a frame image through a camera; transmitting, to a server, a first global pose of the camera and the generated frame image; and receiving, from the server, a second global pose of the camera estimated on the basis of a pose of an object included in the transmitted frame image.

Abstract: A method for normalizing platform-adaptive audio includes encoding input video content and generating video stream data as original data to store the video stream data in storage; generating loudness metadata for audio data of the video content and storing the loudness metadata in the storage; receiving a request for the video content from a client; searching the storage for video stream data of the video content corresponding to the request, the loudness metadata, and a device profile corresponding to device information included in the request; and transmitting, to the client, a response including the video stream data, the loudness metadata, and the device profile that are found in the storage.

Abstract: A robot joint device having an upper arm unit including a first rotating pulley, a second rotating pulley and a third rotating pulley sequentially arranged in one direction, a forearm unit including a fixing pulley configured to rolling-contact the third rotating pulley, a connection unit including a first pivot shaft passing through the third rotating pulley and being connected to the upper arm unit, and a second pivot shaft passing through the fixing pulley and being connected to the forearm unit, a first cable unit wound around the first rotating pulley in a first direction, bent between the second rotating pulley and the third rotating pulley, and fixed to the fixing pulley, and a second cable unit wound around the first rotating pulley in a second direction opposite to the first direction, bent between the second rotating pulley and the third rotating pulley, and fixed to the fixing pulley.

Abstract: A robot joint device may include a first frame unit having a first hollow portion, a second frame unit having a second hollow portion, the second frame unit spaced apart from the first frame unit, a rotational axis unit on and between the first and second hollow portions and configured to rotate around first, second and third axes which are perpendicular to one another, a first link unit extending to bypass the rotational axis unit, connected to each of the first and second frame units with the rotational axis unit interposed therebetween, and configured to rotate around the first and second axes, and a second link unit extending to bypass the rotational axis unit while intersecting with the first link unit, connected to each of the first and second frame units with the rotational axis unit interposed therebetween, and configured to rotate around the first and second axes.

Abstract: A training system includes: an iterative attention module configured to, based on first features in input images, determine ordered sets of second features using iterative attention over T iterations, where T is an integer greater than or equal to two; and a training module configured to: selectively input pairs of matching images to the iterative attention module; selectively input non-matching images to the iterative attention module; and based on the ordered sets generated by the iterative attention module based on the input pairs of matching images and the input non-matching images, train the iterative attention module based on minimizing at least one of: a contrastive loss; and a cosine similarity loss.

Abstract: A method for implementing a seamless switching mode between channels in a multi-stream live transmission environment includes receiving in a single stream, a composite image in which images of multiple channels are synthesized into a single image, composing a view mode having a layout including the images of the multiple channels using the composite image, and changing a layout of the view mode using the composite image.

Abstract: The present invention relates to methods for learning a robot task and robots systems using the same. A robot system may include a robot configured to perform a task, and detect force information related to the task, a haptic controller configured to be manipulatable for teaching the robot, the haptic controller configured to output a haptic feedback based on the force information while teaching of the task to the robot is performed, a sensor configured to sense first information related to a task environment of the robot and second information related to a driving state of the robot, while the teaching is performed by the haptic controller for outputting the haptic feedback, and a computer configured to learn a motion of the robot related to the task, by using the first information and the second information, such that the robot autonomously performs the task.

Abstract: A method of jointly training of a transferable feature extractor network, an ordinal regressor network, and an order classifier network in an ordinal regression unsupervised domain adaption network by providing a source of labeled source images and unlabeled target images; outputting image representations from a transferable feature extractor network by performing a minimax optimization procedure on the source of labeled source images and unlabeled target images; training a domain discriminator network, using the image representations from the transferable feature extractor network, to distinguish between source images and target images; training an ordinal regressor network using a full set of source images from the transferable feature extractor network; and training an order classifier network using a full set of source images from said transferable feature extractor network.