Abstract: There is provided a VIL system-based autonomous driving function verification method. According to embodiments of the disclosure, a VIL system enables an autonomous vehicle to verify an autonomous driving function by interlocking with a virtual road environment in any other place, without having to go to a real test road corresponding to a simulated virtual road environment. Accordingly, an autonomous driving function can be rapidly verified based on a VIL system with respect to various virtual road environments without changing a driving place, so that speed and convenience in development of autonomous driving technology can be enhanced.

Abstract: Provided are a method and a system for generating an AI training hierarchical dataset including data acquisition context information. A GT dataset generation method according to an embodiment of the present disclosure includes: acquiring and storing vehicle data; acquiring and storing sensor data generated at a sensor installed in a vehicle; and generating and storing context information which is information regarding a context at a time when the data is acquired. Accordingly, in generating a GT descriptor, various contexts, conditions at the time when data is acquired may be made to be easily analyzed, classified on the GT descriptor through a hierarchical dataset, which hierarchically describes context information at the time when sensor data is acquired on the descriptor, so that an AI network is effectively trained, and eventually, has high recognition performance.

Abstract: Provided is a method for filtering driving paths in order to stably track a path which is generated for various purposes, such as recognizing a lane and keeping or changing the lane, or avoiding an obstacle around a road. A driving path filtering system according to an embodiment of the present disclosure includes: a position recognition unit configured to recognize a position of a vehicle; a lane recognition unit configured to recognize a lane of a road; and a path generator configured to generate a path for the vehicle to travel on, based on a result of recognizing the position and a result of recognizing the lane, and to perform path noise filtering with respect to the generated path. Accordingly, a driving path can be stabilized by removing a path noise (error) by complementally filtering a lane recognition-based path generation method and a map-based path generation method according to a driving environment.

Abstract: Provided is a deep learning-based panoptic segmentation accelerated processing technique using a complexity-based RPN skip method. An image segmentation system includes: a first processing unit configured to extract dynamic objects in an instance segmentation method by using an extracted feature; a calculation unit configured to control to skip some areas of the feature extracted at the network by the first processing unit, on the basis of complexity of the input image; a second processing unit configured to extract static objects in a semantic segmentation method by using the feature extracted at the network; and a fusion unit configured to fuse a result of extracting by the first processing unit and a result of extracting by the second processing unit. Accordingly, the panoptic segmentation method can be easily performed even in an embedded environment by reducing complexity for panoptic segmentation processing by reducing a calculation burden.

Abstract: There are provided an apparatus and a method for LiDAR data conversion for training various types of autonomous vehicles by using pre-acquired data. A method for converting LiDAR data according to an embodiment includes: receiving an input of first LiDAR data which is pre-acquired through a first LiDAR sensor mounted in a first vehicle; converting the inputted first LiDAR data into second LiDAR data which is acquired through a second LiDAR sensor mounted in a second vehicle; and outputting the converted second LiDAR data, and converting includes converting the first LiDAR data into LiDAR data on a reference coordinate system, and converting the converted LiDAR data into the second LiDAR data which is LiDAR data on a coordinate system of the second LiDAR sensor.

Abstract: There are provided a behavior recognition method for efficient recognition of hand signals and gesture and a behavior recognition AI network system using the same. The behavior recognition method for efficient recognition of hand signals and gestures according to an embodiment includes: an input feature extraction step of extracting, by a behavior recognition AI network system, key points F1 from bounding box data of an object which makes hand signals to be inputted by sequence, and generating skeleton data of the object; and a spatial feature extraction step of calculating, by the behavior recognition AI network system, a length F2 and an angle F3 of a bone vector based on the key points F1 and the skeleton data, and extracting a spatial feature. Accordingly, performance of recognition of hand signals and gestures may be enhanced.

Abstract: Provided is a traffic simulation for controlling a motion of an object, such as a vehicle, a pedestrian moving on a road or a pavement, in a driving simulation, an autonomous driving simulation, or the like. A traffic simulation method according to an embodiment of the present disclosure includes the steps of: importing a new moving object into a simulation environment of a simulator; retrieving data of a moving path and a start point of the moving object which is created based on a function, among pre-stored data; calculating 3D coordinates regarding a position of the moving object; moving the moving object along the moving path in the simulation environment, based on the calculated 3D coordinates; and calculating a next position of the moving object. Accordingly, a motion of an object within a simulator may be precisely created and reliability of validation regarding an operation of an algorithm mounted in an autonomous driving vehicle may be enhanced.

Abstract: Provided is an air housing apparatus for protecting, from contamination, scratches, damage and the like, a lens and a cover glass which affect the performance of an optical device installed on a vehicle such as a camera and LIDAR. The air housing apparatus according to an embodiment of the present invention comprises: a cover into which high pressure air is fed; an air guide for injecting the supplied high pressure air toward the front surface of a lens or a sensor; and an air housing provided so as to connect the cover and the air guide and transferring the high pressure air supplied from the cover to the air guide. Therefore, a lens, sensor, or cover glass of a vehicle-installed optical device such as a camera and LIDAR can be continually protected from contamination and scratches by means of the injected air.

Abstract: A stereo vision SoC and a processing method thereof are provided. The stereo vision SoC extracts first support points from an image and adds second support points, performs triangulation based on the first support points and the second support points; and extracts disparity using a result of the triangulation. Accordingly, depth image quality is improved and HW is easily implemented in the stereo vision SoC.

Abstract: A stereo vision SoC and a processing method thereof are provided. The stereo vision SoC extracts first support points from an image and adds second support points, performs triangulation based on the first support points and the second support points; and extracts disparity using a result of the triangulation. Accordingly, depth image quality is improved and HW is easily implemented in the stereo vision SoC.

Abstract: A glasses type device with an operating means on a rim and a user input method using the same are provided. The glasses type device includes: a touch sensor provided on a rim of a frame of a glasses shape; and a processor configured to process a user operation which is input through the touch sensor. Accordingly, the touch sensors for inputting a user operation are located on the rim of the glasses frame for fixing lenses, so that intuitive various user operations can be performed.

Abstract: Provided are an apparatus and method for providing augmented reality information using a glasses-type wearable device or head mount display. The present invention may provide augmented reality information generated by overlaying POI information in a gaze direction of a user on an image in the direction. Also, the present invention may compare main POI with an object in the image to adjust position information of the user when providing the augmented reality information, thereby providing the augmented reality information having high visibility and accuracy, with the size of UI being adjusted according to the distance from the user. The present invention may allow the user to designate an object in the image by a 3D pointing operation of the user, thereby displaying more precise augmented reality information.

Abstract: Provided is a black box timing modeling method for a digital circuit comprising synchronous elements including latches. The method includes: characterizing a setup time arc by extracting a setup time with respect to a rising or falling edge of a clock of a synchronous element with respect to an input connected to the synchronous element and forming the setup time arc using the extracted setup time; and characterizing a clock-to-output delay arc by providing information on an output departure time from an output based on a rising or falling edge of a clock of a closest synchronous element connected to the output, at least partially based on the setup time arc and forming the clock-to-output delay arc. Accordingly, the method can be efficiently used for a latch-based design without re-verifying internal components of the latch-based design during an upper-level verification, thereby reducing verification time and model size.

Abstract: Provided is a black box timing modeling method for a digital circuit comprising synchronous elements including latches. The method includes: characterizing a setup time arc by extracting a setup time with respect to a rising or falling edge of a clock of a synchronous element with respect to an input connected to the synchronous element and forming the setup time arc using the extracted setup time; and characterizing a clock-to-output delay arc by providing information on an output departure time from an output based on a rising or falling edge of a clock of a closest synchronous element connected to the output, at least partially based on the setup time arc and forming the clock-to-output delay arc. Accordingly, the method can be efficiently used for a latch-based design without re-verifying internal components of the latch-based design during an upper-level verification, thereby reducing verification time and model size.