Electronics and Telecommunications Research Institute

Jeong Si Kim

Seung Min Park

Hyun Woo Cho

Chae Deok Lim

Chan Su Yu

Dong Soo Han

A method for deep metric learning per hierarchical step of multi-labels and few-shot inference using the same is provided, which includes sampling a plurality of N multiplets from a learning data set that is expressed by hierarchical multi-labels; configuring the N multiplets into triplets or quadruplets so as to correspond to the number of classes in each hierarchy; performing a metric learning per hierarchical step for a deep neural network model based on the configured triplets and quadruplets; and distributing, to an embedded device, the deep neural network model from which a loss function is removed as the metric learning is completed.

Method, server, and system for deep metric learning per hierarchical steps of multi-labels and few-shot inference using the same

Disclosed are a method of providing an embedded software development tool for supporting an embedded software development environment based on native building, and an apparatus thereof. The method includes extracting information about a toolchain according to a request by the host system, and providing the extracted information to the host system, generating a project execution file by receiving a project, to which the information about the toolchain is applied, from the host system, and building the project, and generating a project execution result by executing the project execution file, thereby providing an embedded software development environment capable of minimizing the economic costs incurred for acquiring a cross toolchain corresponding to hardware and a platform of a target system and performing a development process when embedded software is developed.

Method of providing embedded software development tool for supporting embedded software development environment based on native building, and apparatus thereof

Provided are a tool for supporting open computing language (OpenCL) application software development for an embedded system and a supporting method thereof. The tool in conjunction with a target system includes a project management unit configured to separate and store OpenCL application software constituted of a host program and a kernel program into a host source code corresponding to the host program and a kernel source code corresponding to the kernel program, and manage a tool-chain corresponding to an OpenCL platform of the target system, and a builder configured to determine a compile type of the kernel source code and determine whether the kernel source code is built in response to the compile type. Accordingly, it is possible to rapidly provide a tool for supporting OpenCL application software development that can develop OpenCL application software for an embedded system.

Tool composition for supporting openCL application software development for embedded system and method thereof

A host system having a device driver development tool and a target system having a device driver agent, the tool and the target system separated from each other and interconnected with a communication network. The driver agent device communicates with the target system, receiving and processing various service requests from the device driver development tool, and transmits the processed services to the host system through the communication network. Thr driver agent determines types of the services requested from the device driver development tool and performs services based on the types of the services determined. Accordingly, the device driver can be adapted to the target system and tested without complex procedures to effectively reduce time and manpower needed for the device driver development.

Driver agent device for supporting remote device driver development environment in embedded system and operation method thereof

A parallel loop transformation method for race detection during an execution of parallel programs that includes generating a data structure of a condition statement branch determinant string Cstr required for loop transformation by taking an original parallel loop as an input and extracting execution path information, transforming the original parallel loop into a full race covering loop using the data structure of the condition statement branch determinant string Cstr required for loop transformation and the execution path information as an input statement, instrumenting the race detection function in order to activate the race detection function for the transformed parallel loop which are previously generated, and executing the race detection while running the parallel programs according to instrumented detection functions.

Growing community of inventors

Daejeon, South Korea

Jeong Si Kim