Abstract: An image processing method, device, storage medium and camera are provided. The method, applied to the camera, comprises: capturing a target image; acquiring a target feature map of the target image through a preset target convolution layer, wherein the target convolution layer includes at least one of a plurality of convolution layers of a convolutional neural network (CNN); and outputting the target feature map. That is to say, after the target image is captured by the camera, the target feature map may be acquired by processing the target image through the target convolution layer pre-integrated in the camera. In this way, the camera transmits the target feature map only to reduce the transmitted data volume, thereby being capable of shortening transmission delay and saving bandwidth required by image transmission.

Abstract: A method for tracking target object, storage medium and electronic device, which relate to the field of an image processing technology. The method includes: receiving at least one target image captured by a single photon avalanche diode (SPAD) camera before present moment; for each target image, inputting the target image and a preset template image into a pre-trained siamese network to acquire a position of a target object in the target image output by the siamese network, wherein the template image includes the target object; and determining a position of the target object in an image to be predicted based on the position of the target object in each target image, wherein the image to be predicted is an image captured by the SPAD camera at the present moment.

Abstract: The present disclosure relates to a target tracking method, device, medium and apparatus, belongs to the technical field of computers, and can ensure the completeness of tracking. The target tracking method comprises performing object detection on all telephoto image frames captured by telephoto cameras in a multi-camera system, and constructing a detection point set based on object detection result for each frame of the telephoto image; selecting a target tracking point from each detection point set based on the starting tracking point; and connecting the selected target tracking points as a tracking sequence for the starting tracking point.

Abstract: A method for determining an academic impact, including: pre-processing a citing paper to obtain pre-processed data; determining a category of an impact, on the citing paper, of each reference paper and ranking of contribution, to the citing paper, of each reference paper, according to the pre-processed data; determining a local influential factor, on the citing paper, of each reference paper, according to the category of the impact and the ranking of the contribution of each reference paper; and determining a global influential factor of each reference paper according to a global influential factor of the citing paper and the local influential factor.

Abstract: The present disclosure relates to a method, apparatus, electronic device, storage medium and system for vision task execution, and belongs to the technical field of information. The method includes: acquiring raw image data which is collected by an image sensor and not processed by an image signal processor (ISP); and inputting the raw image data into a trained first raw image neural network for executing a target vision task to obtain a vision task processing result output by the first raw image neural network, wherein the first raw image neural network is obtained based on training with an raw image training set. A vision task in response is executed by employing the raw image data so as to solve the problems that the quantity of image data processed by the ISP is relatively large, and related information may be lost in the related art.

Abstract: The present disclosure relates to a target tracking method, device, medium and apparatus, belongs to the technical field of computers, and can ensure the completeness of tracking. The target tracking method comprises performing object detection on all telephoto image frames captured by telephoto cameras in a multi-camera system, and constructing a detection point set based on object detection result for each frame of the telephoto image; selecting a target tracking point from each detection point set based on the starting tracking point; and connecting the selected target tracking points as a tracking sequence for the starting tracking point.

Abstract: A video processing method, including: acquiring a plurality of raw videos, the plurality of raw videos being videos acquired by a plurality of dynamic image recorders arranged according to preset locations; determining an overlapping region between every two adjacent raw videos according to preset rules corresponding to the preset locations, the adjacent raw videos being the raw videos acquired by the dynamic image recorders arranged adjacent to each other; performing multi-stage optical flow calculation on the raw videos in each overlapping region to obtain a plurality of pieces of target optical flow information; and splicing the overlapping region of every two adjacent raw videos based on the target optical flow information to obtain target videos.

Abstract: The present disclosure relates to a focus tracking system and method.

Abstract: The present disclosure relates to a focus tracking system and method.

Abstract: The present disclosure relates to a method, apparatus, electronic device, storage medium and system for vision task execution, and belongs to the technical field of information. The method includes: acquiring raw image data which is collected by an image sensor and not processed by an image signal processor (ISP); and inputting the raw image data into a trained first raw image neural network for executing a target vision task to obtain a vision task processing result output by the first raw image neural network, wherein the first raw image neural network is obtained based on training with an raw image training set. A vision task in response is executed by employing the raw image data so as to solve the problems that the quantity of image data processed by the ISP is relatively large, and related information may be lost in the related art.

Abstract: The present disclosure relates to a method and a system for joint optimization of an ISP and vision tasks, a medium and an electronic device, which belong to the field of image processing and can effectively avoid the over-fitting of joint optimization of the ISP and the vision tasks. The method for joint optimization of the ISP and the vision tasks includes the following steps: performing image signal processing on raw image dataset by an ISP to obtain processed image dataset; measuring probability gradient of the processed image dataset in prior distribution of traditional image dataset by a measurement module; and performing vision tasks on the processed image dataset by using a loss function with the probability gradient as a regularization term via a neural network.

Abstract: An image processing method, device, storage medium and camera are provided. The method, applied to the camera, comprises: capturing a target image; acquiring a target feature map of the target image through a preset target convolution layer, wherein the target convolution layer includes at least one of a plurality of convolution layers of a convolutional neural network (CNN); and outputting the target feature map. That is to say, after the target image is captured by the camera, the target feature map may be acquired by processing the target image through the target convolution layer pre-integrated in the camera. In this way, the camera transmits the target feature map only to reduce the transmitted data volume, thereby being capable of shortening transmission delay and saving bandwidth required by image transmission.

Abstract: A method for tracking target object, storage medium and electronic device, which relate to the field of an image processing technology. The method includes: receiving at least one target image captured by a single photon avalanche diode (SPAD) camera before present moment; for each target image, inputting the target image and a preset template image into a pre-trained siamese network to acquire a position of a target object in the target image output by the siamese network, wherein the template image includes the target object; and determining a position of the target object in an image to be predicted based on the position of the target object in each target image, wherein the image to be predicted is an image captured by the SPAD camera at the present moment.

Abstract: The present invention discloses an optimized coding method for an omnidirectional video, computer readable storage medium and computer device to solve the technical problem that the video quality cannot be guaranteed under a low code rate in the prior art. The method includes: obtaining attribute information of each coding unit of an omnidirectional video file, wherein each coding unit is a storage and coding unit of the omnidirectional video file; determining a coding mode corresponding to each coding unit according to the attribute information of each coding unit; and coding each coding unit according to the coding mode corresponding to each coding unit.

Abstract: The present invention discloses an optimized coding method for an omnidirectional video, computer readable storage medium and computer device to solve the technical problem that the video quality cannot be guaranteed under a low code rate in the prior art. The method includes: obtaining attribute information of each coding unit of an omnidirectional video file, wherein each coding unit is a storage and coding unit of the omnidirectional video file; determining a coding mode corresponding to each coding unit according to the attribute information of each coding unit; and coding each coding unit according to the coding mode corresponding to each coding unit.

Abstract: A congestion control mechanism for TCP communication sessions is described. The congestion control mechanism adjusts the size of the congestion window based on a number of parallel virtual TCP Hybla connections. The number of parallel virtual TCP Hybla connections used to determine the congestion window can be dynamically adjusted. The enhanced TCP Hybla congestion control mechanism provides improved throughput for heterogenous networks having a high latency link and having possible packet loss.

Abstract: A congestion control mechanism for TCP communication sessions is described. The congestion control mechanism adjusts the size of the congestion window based on a number, N, of parallel virtual connections. The number N of parallel virtual connections used to determine the congestion window is dynamically adjusted based on an estimation of the network condition.

Abstract: A congestion control mechanism for TCP communication sessions is described. The congestion control mechanism adjusts the size of the congestion window based on a number of parallel virtual TCP Hybla connections. The number of parallel virtual TCP Hybla connections used to determine the congestion window can be dynamically adjusted. The enhanced TCP Hybla congestion control mechanism provides improved throughput for heterogenous networks having a high latency link and having possible packet loss.

Abstract: A congestion control mechanism for TCP communication sessions is described. The congestion control mechanism adjusts the size of the congestion window based on a number, N, of parallel virtual connections. The number N of parallel virtual connections used to determine the congestion window is dynamically adjusted based on an estimation of the network condition.