Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating a depth image, comprising obtaining data representing a first image generated by a first sensor and a second image generated by a second sensor, wherein each of the first and second images includes a plurality of pixels; determining, for each pixel of the plurality of pixels included in the first image, whether the pixel is a boundary pixel associated with a boundary of an object that is represented in the first image; determining, from a plurality of candidate penalty values and for each pixel in the first image, an optimized penalty value for the pixel; generating an optimized cost function for the first image based on the optimized penalty values for the plurality of pixels; and generating a depth image for the first image based on the optimized cost function.

Abstract: The present invention discloses system and method for enhancing an image based on non-local features. The invention includes an end-to-end trainable and guided method including feature extraction block, non-local feature generator and non-local feature enhancement block, to deal with low-level image problem by using the non-local feature concept. The invention also deploys a non-local feature merge block to rectify translated features and improve non-local feature further to finally reconstruct the rectified features to form an enhanced image.

Abstract: A method of constructing an adaptive multiply accumulate layer in a convolutional neural network, including determining an activation data map width, an activation data map height, a channel depth, a batch, a kernel width, a kernel height and a filter set number, setting a first dimension of an adaptive multiplier layer based on the activation data map width, setting a second dimension of the adaptive multiplier layer based on the channel depth, setting a third dimension of the adaptive multiplier layer based on the filter set number and constructing the adaptive multiplier layer based on the first dimension, the second dimension and the third dimension.

Abstract: The present disclosure relates to a method for inter-core communication of a multi-core processor, a processor, an inter-core communication system and a computer readable storage medium. The method for inter-core communication comprises: receiving a communication request sent by an initiating core for communication with receiving cores; and instructing, on the basis of the communication request, the initiating core to communicate with the receiving cores by using a delivery message to invoke interfaces of services in the receiving cores; wherein the delivery message is service-oriented and corresponds to the interfaces of the services in the receiving cores. According to the method for inter-core communication of the present disclosure, services built in various systems can interact in a uniform and universal manner by means of the service-oriented delivery message. Therefore, the portability of application programs located on different cores can be improved, and the development difficulty is reduced.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for controlling, by an on-chip memory controller, a plurality of hardware components that are configured to perform computations to access a shared memory. One of the on-chip memory controller includes at least one backside arbitration controller communicatively coupled with a memory bank group and a first hardware component, wherein the at least one backside arbitration controller is configured to perform bus arbitrations to determine whether the first hardware component can access the memory bank group using a first memory access protocol; and a frontside arbitration controller communicatively coupled with the memory bank group and a second hardware component, wherein the frontside arbitration controller is configured to perform bus arbitrations to determine whether the second hardware component can access the memory bank group using a second memory access protocol different from the first memory access protocol.

Abstract: The current disclosure relates to a panoramic look-around view generation method, an in-vehicle device and an in-vehicle system.

Abstract: A method of describing a temporal event, including receiving a video sequence of the temporal event, extracting at least one physical characteristic of an at least one occupant within the video sequence, extracting at least one action of the at least one occupant within the video sequence, extracting at least one interaction of the at least one occupant with a secondary occupant within the video sequence, determining a safety level of the temporal event within a vehicle based on at least one of the at least one action and the at least one interaction and describing the at least one physical characteristic of the at least one occupant and at least one of the at least one action and the at least one interaction of the at least one occupant.

Abstract: A time synchronization method, apparatus, domain controller, and storage medium are disclosed. The domain controller is mounted in a vehicle and includes a number of SoCs and micro control units. The SoCs and micro control units are respectively, communicatively connected through the controller area network bus and respectively connected to the switch via Ethernet. The switch has external Ethernet interfaces. A main SoC in the number of SoCs has external UART/PPS interfaces and the micro control units have external FlexRay interfaces. The time synchronization method of the domain controller includes any one of the steps of receiving a time service from an external device through the UART/PPS interfaces, receiving a time service from the external device through the switch over the Ethernet interface, or receiving a time service from the external device through the FlexRay interface under a normal operation phase of the vehicle.

Abstract: A sensor configuration method, an apparatus, computer equipment and a storage medium are disclosed. The method includes acquiring connection configuration information generated based on user input, querying for a target sensor driver matched with a target sensor represented by the sensor identifier in a device driver set, calling a target communication port driver matched with the target communication port represented by the communication port identifier and establishing communications with the target sensor by the target sensor driver and the target communication port driver. The connection configuration information includes a sensor identifier and a communication port identifier. The device driver set is pre-stored with a number of sensor drivers corresponding to a number of sensors.

Abstract: The present disclosure provides a reverse assist method for a vehicle, comprising: transforming an i-th image in a reversing direction acquired by a camera unit at time Ti into an i-th top plan view, wherein i?1; generating an i-th result image based on the i-th top plan view, wherein when i is greater than 1, the i-th top plan view and an intermediate image are spliced to generate an i-th result image; outputting the i-th result image; and wherein the intermediate image is a view transformed from an image of an outside environment in the reversing direction acquired at a time earlier than Ti, and the intermediate image is updated according to a predetermined condition during reversing.

Abstract: An example method of single-exposure multiple-frame image capture, making an HDR video, comprising, capturing a series of frames at an underexposed setting, selecting a reference image from the series of frames, aligning the series of frames to the reference image, merging the series of frames into one frame based on the reference image, indexing the reference image to a next frame in the series of frames, resulting in an indexed reference image, realigning the series of frames to the indexed reference image and remerging the series of frames into a single frame based on the indexed reference image.

Abstract: An anti-shake image processing method, apparatus, electronic device and storage medium are disclosed. The method includes the following steps. A current actual posture of an imaging device at a current shooting moment of capturing an original image of current frame and multiple reference actual postures of the imaging device at multiple shooting moments of capturing the original images of adjacent multiple frames of the original image of the current frame are obtained. A path smoothing process is performed to determine a current virtual posture after path smoothing at the current shooting moment. A coordinate transformation is performed on the original image of the current frame captured in the current actual posture to an estimated position of the original image of the current frame when captured in the current virtual posture in a pixel coordinate system to obtain a first correction image of the current frame.

Abstract: The present application relates to a kernel debugging system and method. The kernel debugging system includes: a user interface module, the user interface module being configured to edit program codes and output an execution result of the program codes; a compilation module, the compilation module being configured to compile the program codes into object files; and a Kwasm engine, the Kwasm engine being directly installed in a system kernel of an operating system, and being configured to interpret and execute the object files in a kernel mode, so as to obtain the execution result of the program codes. On the basis of the kernel debugging system, a user can write program codes like writing a common application program, namely, the program codes can be run and executed in a system kernel without paying attention to details of the system kernel.

Abstract: An extensible multi-precision data pipeline system, comprising, a local buffer that stores an input local data set in a local storage format, an input tensor shaper coupled to the local buffer that reads the input local data set and converts the input local data set into an input tensor data set having a tensor format of vector width N by tensor length L, a cascaded pipeline coupled to the input tensor shaper that routes the input tensor data set through at least one function stage resulting in an output tensor data set, an output tensor shaper coupled to the cascaded pipeline that converts the output tensor data set into an output local data set having the local storage format and wherein the output tensor shaper writes the output local data set to the local buffer.

Abstract: The present invention discloses a system and method for image processing and recognizing a scene of an image. The system utilizes a Multi-mode scalable network system and regrouping pipeline. The system is AI based system which uses neuro network. The system includes a pre-processing, processing and a post-processing unit. The system uses optical information recorded from the camera of a mobile device to extract and analyze the content in an image such as a photo or video clip. Based on the retrieved information, a label is given to best describe the scene of the image.

Abstract: A method of instance segmentation, comprising, defining a bounding box, depth mapping the bounding box, sorting the depth map, estimating stability due to noise of the sorted depth map, detecting low frequency jumps of the sorted depth map based on the estimated stability, detecting high frequency jumps of the sorted depth map, comparing the detected low frequency jumps and the detected high frequency jumps, detecting at least one of a largest object and a closest object based on the comparison and pixel masking the at least one of the largest object and the closest object.

Abstract: An example operation of depth map generation includes one or more of, simultaneously capturing a main-off camera image and an auxiliary-off camera image with an unpowered flash, sparse depth mapping an object based on the main-off camera image and the auxiliary-off camera image, capturing a main-on camera image with a powered flash, foreground probability mapping the object based on the main-off camera image and the main-on camera image and dense depth mapping the object based on the sparse depth map and the foreground probability map.

Abstract: The present invention discloses system and method for processing an image. The invention processes the image by segmenting a human portrait region of the image. The invention uses ahierarchical hybrid loss module for masking the portrait region generating masked portrait region. The invention also uses data learning the masked portrait region.

Abstract: The present invention discloses a system for precise representation of object segmentation with multi-modal input for real-time video applications. The multi-modal segmentation system takes advantage of optical, temporal as well as spatial information to enhance the segmentation for AR and VR or other entrainment purpose with accurate details. The system can segment foreground objects such as human and salient objects within a video frame and allows locating object-of-interest for multiple-purposes.

Abstract: A method of motion control and textured display, comprising, receiving an image, extracting an initial mask of a subject of interest from the image, segmenting the image into the initial mask and a background, determining a color texture of the initial mask, estimating motion of the initial mask to a new mask and merging the new mask and the background into a merged image.