Abstract:
Provided is a neural network accelerator which performs a calculation of a neural network provided with layers, the neural network accelerator including a kernel memory configured to store kernel data related to a filter, a feature map memory configured to store feature map data which are outputs of the layers, and a Processing Element (PE) array including PEs arranged along first and second directions, wherein each of the PEs performs a calculation using the feature map data transmitted in the first direction from the feature map memory and the kernel data transmitted in the second direction from the kernel memory, and transmits a calculation result to the feature map memory in a third direction opposite to the first direction.
Abstract:
Provided is a sensing circuit for recognizing a movement including: at least one light emitting device outputting light; at least one light receiving device receiving the light reflected by an object on the light emitting device and generating a plurality of current signals proportional to an amount of incident light; a signal conversion unit converting the plurality of current signals into a plurality of digital signals; a recognition unit measuring a synthetic digital signal to determine whether an object moves by receiving the plurality of current signals; and a control unit controlling the recognition unit, wherein the recognition unit generates a clock signal for the synthetic digital signal greater than a critical value and measures a count generated by the clock signal; and the control unit determines whether the object moves through a comparison of the count and a reference value.
Abstract:
A convolutional operation device for performing convolutional neural network processing includes an input sharing network including first and second input feature map registers configured to shift each input feature map, which is inputted in row units, in a row or column direction and output the shifted input feature map and arranged in rows and columns, a first MAC array connected to the first input feature map registers, an input feature map switching network configured to select one of the first and second input feature map registers, a second MAC array connected to one selected by the input feature map switching network among the first and second input feature map registers, and an output shift network configured to shift the output feature map from the first MAC array and the second MAC array to transmit the shifted output feature map to an output memory.
Abstract:
The present disclosure relates to a frame grabber, an image processing system, and an image processing method. A frame grabber according to an embodiment of the inventive concept includes a plurality of decoders, a plurality of image controllers, a plurality of memories, a synchronization controller, and a synchronization memory. The plurality of decoders generate a plurality of image data by decoding a plurality of image signals. The plurality of image controllers generate a plurality of pixel data and a plurality of frame information data on the basis of the plurality of image data. The plurality of memories store the plurality of pixel data. The synchronization controller receives the plurality of frame information data, and generates synchronization data on the basis of the plurality of frame information data. The synchronization memory stores the frame information data and the synchronization data.
Abstract:
A differential driving circuit according to embodiments of the inventive may include a first driver drives a first pad to a first voltage according to a first driving signal, a second driver drives a second pad to a second voltage according to a second driving signal, a first and second capacitors for receiving a first and second voltage changes of the first and the second pad at one end thereof respectively to transmit the first and the second voltage change to the other end thereof respectively in a transition interval in which voltages of the first and second pads are changed, transition interval voltage adder circuit adds voltages respectively transmitted thereto through the first and second capacitors, and a transition interval asymmetry compensation circuit adjusts a slope of at least one of the first and second driving signals according to the added voltage.
Abstract:
Disclosed is an operating method of a vehicle control apparatus controlling autonomous driving based on a vehicle external object including performing primary object detection based on a first vehicle external image received from a camera to obtain first object information, setting a first reflective area for reflection light based on the first object information, generating a second vehicle external image, in which a reflective image inside the first reflective area is removed from the first vehicle external image, using pixel values inside the first reflective area, performing secondary object detection based on the second vehicle external image to obtain second object information, determining reliability of the second object information based on information about the reflective image and the second object information, and controlling the autonomous driving of the vehicle based on the second object information when the reliability of the second object information is higher than a setting value.
Abstract:
A reorganizable neural network computing device is provided. The computing device includes a data processing array unit including a plurality of operators disposed at locations corresponding to a row and a column. One or more chaining paths which transfer the first input data from the operator of the first row of the data processing array to the operator of the second row are optionally formed. The plurality of first data input processors of the computing device transfer the first input data for a layer of the neural network to the operators along rows of the data processing array unit, and the plurality of second data input processors of the computing device transfer the second input data to the operators along the columns of the data processing array.
Abstract:
Disclosed is a parallel processor. The parallel processor includes a processing element array including a plurality of processing elements arranged in rows and columns, a row memory group including row memories corresponding to rows of the processing elements, a column memory group including column memories corresponding to columns of the processing elements, and a controller to generate a first address and a second address, to send the first address to the row memory group, and to send the second address to the column memory group. The controller supports convolution operations having mutually different forms, by changing a scheme of generating the first address.
Abstract:
Provided is a data generation device for generating input data to be inputted to a parallel processing device. The data generation device includes: a controller configured to output padding data; and a data processing device configured to receive original data and to generate the input data in which at least a portion of the original data is padded with the padding data. The data processing device includes: a first multiplexer configured to receive the padding data and the original data; a register configured to store data outputted from the first multiplexer; and a second multiplexer configured to receive data outputted from the first multiplexer and data stored in the register.
Abstract:
Provided are an object recognition device, an autonomous driving system including the same, and an object recognition method using the object recognition device. The object recognition device includes an object frame information generation unit, a frame analysis unit, an object priority calculator, a frame complexity calculator, and a mode control unit. The object frame information generation unit generates object frame information based on a mode control signal. The frame analysis unit generates object tracking information based on object frame information. The object priority calculator generates based on object tracking information. The frame complexity calculator generates a frame complexity based on object tracking information. The mode control unit generates a mode control signal for adjusting an object recognition range and a calculation amount of the object frame information generation unit based on the priority information, the frame complexity, and the resource occupation state.