Abstract: Systems, devices, and methods for performing feature compression, including obtaining an input video; obtaining a feature map corresponding to the input video, the feature map including a plurality of feature channels; reordering the plurality of feature channels based on at least one characteristic of the plurality of feature channels; compressing the reordered plurality of feature channels; and generating an encoded bitstream based on the compressed and reordered plurality of feature channels.

Abstract: A production line for producing components to a high standard of cleanliness and sealed and protected in that state includes a loading device, a cleaning device, a detecting device, a pasting device, a heat-sealing device, a packing device, and transfer devices of the production line. The production line automatically processes the components for obtaining components with the high cleanliness. By the processes of protective film pasting, heat-sealing, and packing, the components may be further protected from subsequent pollution. A method for producing components with a high cleanliness applied to the production line is also disclosed.

Abstract: Methods and apparatuses of encoding a video stream using video point cloud coding include obtaining a source point cloud; obtaining a geometry-reconstructed point cloud, wherein the geometry-reconstructed point cloud is reconstructed using lossy coding; obtaining a target point included in the geometry-reconstructed point cloud; performing at least one of a forward K-dimensional (KD)-tree search to determine a first plurality of points of the source point cloud which are nearest neighbors to the target point, and a backward KD-tree search to determine a second plurality of points of the source point cloud for which the target point is a nearest neighbor; determining an average color value based on at least one of a first average color value of the first plurality of points and a second average color value of the second plurality of points; selecting a color value for the target point based on the average color value; and generating an encoded video stream based on the selected color value.

Abstract: Provided in the present application are an end-edge-cloud coordination system and method based on a digital retina, and a device. The coordination system may include, but is not limited to, a front-end device, an edge device, and a cloud device. The front-end device is used for extracting, from collected video data, features having universality, and is used for generating analysis and recognition tasks on the basis of the features. The front-end device is also used for processing the analysis and recognition tasks, so as to obtain a first intermediate result to be sent to the edge device. The edge device is used for processing the analysis and recognition tasks on the basis of the first intermediate result, so as to obtain a second intermediate result to be sent to the cloud device. The cloud device is used for processing the analysis and recognition tasks on the basis of the second intermediate result, so as to obtain an analysis and recognition result for video data.

Abstract: A method, computer program, and computer system is provided for decoding point cloud data. Data corresponding to a point cloud is received. A number of contexts associated with the received data is reduced based on occupancy data corresponding to one or more parent nodes and one or more child nodes within the received data. The data corresponding to the point cloud is decoded based on the reduced number of contexts.

Abstract: Systems, devices, and methods for fixed point implementation of a subdivision processing for video decoding are provided, which may include obtaining two vertices associated with an encoded mesh, determining a first normal vector and a second normal vector with integer components. It may also include determining a first normalized normal vector based on the first normal vector and a second normalized normal vector based on the second normal vector using an inverse square root function and determining a middle normalized normal vector associated with a middle-point based on the combination of the fix-point number with the first pre-defined number of digits, the right-shift operation by the second pre-defined number, and the components of the middle-point. It may also include decoding the encoded mesh based on the first normalized normal vector, the second normalized normal vector, and the middle normalized normal vector.

Abstract: Provided are a point cloud attribute prediction method and prediction device based on a filter. The method comprises a coding method and a decoding method, and the device comprises a coding device and a decoding device. The method comprises: determining K nearest neighbor points of the current point; determining a filter matrix; and determining an attribute prediction value of the current point according to the filter matrix. Therefore, the compression performance of a point cloud attribute can be improved by means of selecting an appropriate filter.

Abstract: Disclosed is a method and system for achieving optimal separable convolutions, the method is applied to image analyzing and processing and comprises steps of: inputting an image to be analyzed and processed; calculating three sets of parameters of a separable convolution: an internal number of groups, a channel size and a kernel size of each separated convolution, and achieving optimal separable convolution process; and performing deep neural network image process. The method and system in the present disclosure adopts implementation of separable convolution which efficiently reduces a computational complexity of deep neural network process. Comparing to the FFT and low rank approximation approaches, the method and system disclosed in the present disclosure is efficient for both small and large kernel sizes and shall not require a pre-trained model to operate on and can be deployed to applications where resources are highly constrained.

Abstract: A point cloud attribute prediction method and apparatus, a terminal and a storage medium are disclosed. The method includes obtaining target neighbor points corresponding to a target point by a first spatial distance, and determining an optimization weight corresponding to each target neighbor point respectively based on a second spatial distance, and finally determining an attribute prediction value corresponding to the target point based on each target neighbor point and optimization weight corresponding to each target neighbor point respectively. The present disclosure optimizes the weight corresponding to each target neighbor point respectively based on the spatial distance, which can improve the correlation between the geometry information and the attribute information of the point cloud, provide more accurate prediction values when performing point cloud attribute prediction, and thus improve the encoding and decoding performance of the point cloud attributes.

Abstract: A point cloud attribute encoding method and apparatus, a decoding method and apparatus, and a related device are disclosed. The point cloud attribute encoding method includes: sorting all point cloud data to be encoded to acquire sorted point cloud data, the point cloud data to be encoded being point cloud data with attributes to be encoded; grouping the sorted point cloud data based on correlation between the sorted point cloud data to acquire groups to be encoded; and performing point cloud attribute encoding based on the groups to be encoded. It is beneficial for enhancing the correlation between point cloud data within the group, improving the efficiency of decorrelation during intra group transform after grouping, and improving coding efficiency.

Abstract: In a method of point cloud attribute decoding in a point cloud decoder, a sorting method of a set of points of a point is determined. A quantization parameter for quantizing transform coefficients of a Haar-based attribute coding of the point cloud is determined based on first signaling information in a coded bitstream of the point cloud. The Haar-based attribute coding is configured to perform point cloud compression on the set of points of the point cloud based on a Haar transform. An entropy decoding method for decoding the quantized transform coefficients is determined according to the quantization parameter that is determined based on the first signaling information. The quantized transform coefficients are further determined based on the entropy decoding method. Attributes of the point cloud are reconstructed based on the determined sorting method and the determined quantized transform coefficients.

Abstract: A method, computer program, and computer system is provided for decoding point cloud data. Data corresponding to a point cloud is received. A number of contexts associated with the received data is reduced based on reducing a size of an array corresponding to syntax elements for predictive tree-based coding of the point cloud. The data corresponding to the point cloud is decoded based on the reduced number of contexts.

Abstract: A method for deriving motion vector is provided, this method includes: obtaining a space domain motion vector prediction and a time domain motion vector prediction of adjacent blocks of a coding unit in a predetermined direction; performing a filtering operation on the space domain motion vector and the time domain motion vector prediction to obtain the space domain motion vector prediction and the time domain motion vector prediction of the filtered adjacent blocks; determining, according to a predetermined inter-frame prediction mode, reference motion vectors of a current block in four side directions by using the space domain motion vector prediction and the time domain motion vector prediction of the filtered adjacent blocks and a coordinate position of the current block in the coding unit; and deriving motion vectors of the current block according to the reference motion vectors and the coordinate position of the current block in the coding unit.

Abstract: A method of point cloud attribute coding is performed by at least one processor and includes receiving a point cloud data set comprising a bounding box. A determination is made as to whether the bounding box is symmetric or asymmetric. A quad-tree-binary-tree (QtBt) decomposition is performed on the point cloud data set based on the bounding box being asymmetric.

Abstract: Methods and apparatuses of encoding a video stream using video point cloud coding include obtaining a source point cloud; generating a geometry map and a texture map based on the source point cloud; generating a reconstructed geometry map and a reconstructed texture map using lossy coding without using recoloring, wherein a resolution of the reconstructed geometry map is same as a resolution of the reconstructed texture map; obtaining a point of the source point cloud corresponding to a geometry pixel of the geometry map and a texture pixel of the texture map; selecting a color of the texture pixel based on a color of the obtained point; and generating an encoded video stream based on the selected color, the geometry map, and the texture map.

Abstract: A method, computer program, and computer system is provided for point cloud coding. The method includes receiving, from a bitstream, data corresponding to a point cloud; reconstructing, based on the data, a first attribute value of a first duplicate point from among a plurality of duplicate points corresponding to a single geometry position; obtaining at least one prediction residual corresponding to at least one remaining attribute value of at least one remaining duplicate point from among the plurality of duplicate points; reconstructing the at least one remaining attribute value based on the reconstructed first attribute and the at least one prediction residual; and decoding the data corresponding to the point cloud based on the reconstructed first attribute value and the reconstructed at least one remaining attribute value.

Abstract: A technique for encoding video for machine vision and human/machine hybrid vision, including receiving image data. The technique may also include detecting a plurality of bounding boxes associated with a plurality of objects of interest in a frame of the image data and detecting a frame-level bounding box for the frame based on coordinates of the plurality of bounding boxes. Then, the technique may include encoding the frame-level bounding box using a first bitrate.

Abstract: Aspects of the disclosure provide methods and apparatuses for point cloud compression and decompression. In some examples, an apparatus for point cloud compression/decompression includes processing circuitry. In some examples, the processing circuitry receives a bitstream carrying compressed data for a point cloud. The processing circuitry determines that a current node in an octree structure is eligible for an isolated mode. The octree structure corresponds to three dimensional (3D) partitions of a space of the point cloud. Then the processing circuitry determines, based on information of one or more other nodes, a single isolated point flag for the current node that indicates whether the current node is coded with a single isolated point.

Abstract: Embodiments of the present disclosure provide a method and device for selecting a context model of a quantized coefficient end flag. The method comprises: obtaining a scanning position POS of a non-zero coefficient corresponding to current quantized coefficient end flag in a specific scanning order; wherein the scanning position POS is a subscript of the non-zero coefficient in the scanning order; configuring a first context model array, and using a fixed value as the base to calculate the logarithmic value of the value of the scanning position POS plus 1, and according to the logarithmic value, selecting a first context model from the first context model array; and using the first context model to encode or decode a binary symbol of the current quantized coefficient end flag. According to the technical solution of the present application, it is able to improve encoding and decoding efficiency for a quantized coefficient end flag, thereby further improve the efficiency of video encoding and decoding.

Abstract: A method, computer program, and computer system is provided for decoding point cloud data. Data corresponding to a point cloud is received. A number of contexts associated with the received data is reduced based on reducing a size of an array corresponding to syntax elements for predictive tree-based coding of the point cloud. The data corresponding to the point cloud is decoded based on the reduced number of contexts.