Abstract: An apparatus for acquiring an image includes an image sensor, a light compensator, and a light filter assembly. The image sensor is on a light output side of the light filter assembly, and is configured to generate and output a first image signal and a second image signal by a plurality of exposures. The first image signal is generated according to a first preset exposure and the second image signal is generated according to a second preset exposure, and they are two of the plurality of exposures. The light compensator includes a first light compensation apparatus, the first light compensation apparatus is configured to perform near-infrared light compensation, and the near-infrared light compensation is performed in at least part of an exposure period of the first preset exposure but is not performed in an exposure period of the second preset exposure.

Abstract: An anti-counterfeiting face detection method, device and multi-lens camera are disclosed. The anti-counterfeiting face detection method comprises: acquiring a depth image, an infrared image and an RGB image by using a TOF camera and an RGB camera; analyzing the RGB image through a preset face detection algorithm to determine an RGB face region of a face in the RGB image and position information of the RGB face region; determining a depth face region of the face in the depth image and an infrared face region of the face in the infrared image based on the position information of the RGB face region; determining that the face passes the detection when the depth face region, the infrared face region and the RGB face region meet corresponding preset rules respectively.

Abstract: A multi-spectrum based image fusion apparatus is disclosed, which includes a light acquisition device, an image processor, and an image sensor having five types of photosensitive channels. The five types of photosensitive channels including red, green and blue RGB channels, an infrared IR channel and a full-band W channel. The light acquisition device acquires target light corresponding to incident light. The image sensor converts the target light into an image signal through the RGB channels, the IR channel and the W channel. The image processor analyzes the image signal into RGB color signals and a brightness signal, and fuses the RGB color signals and the brightness signal to obtain a fused image. The collection of the channels based on which the RGB color signals and the brightness signal are obtained includes the five types of photosensitive channels.

Abstract: A deep neural network training method and apparatus and a computer device are provided. The deep neural network training method includes: obtaining task attributes of nodes in a current network layer in a tree-like network topology (S101); performing cluster analysis on the nodes in the current network layer based on the task attributes of the nodes in the current network layer and extracting a common part of task attributes of multiple nodes in a same category as a task attribute for a parent node of the multiple nodes (S102); training a network parameter of each parent node based on a task attribute of this parent node (S103); and determining that training of a deep neural network corresponding to the tree-like network topology is completed, after completion of training of all nodes in all network layers (S104). The operation efficiency of deep learning can be improved through this solution.

Abstract: The embodiments of the present disclosure provides a method and an apparatus for training a neural network model, A training sample is obtained, and the neural network model is trained using the training sample. When the neural network model is trained, power exponential domain fixed-point encoding is performed on a first activation inputted into each network layer and a network weight of each network layer, and an encoded first activation and an encoded network weight are power exponential domain fixed-point data, which when used in the operation, can cause a matrix multiplication operation involved to be converted into an addition operation in the power exponential domain by means of the power exponential domain encoding. The hardware resources required for the addition operation are significantly less than that required for the multiplication operation, which therefore can greatly reduce the hardware resource overhead required for running the neural network model.

Abstract: Provided is a system for detecting article store or retrieve operations. The system for detecting store or retrieve operations includes: an article storage-retrieval cabinet, an image acquiring unit, and a storage or retrieval detecting unit, wherein an edge of an entrance-exit of the article storage-retrieval cabinet is provided with a marker, and the image acquiring unit is configured to acquire an image of the entrance-exit of the article storage-retrieval cabinet; and the storage or retrieval detecting unit is connected to the image acquiring unit, and is configured to detect an article store or retrieve operation based on marker information in the image acquired by the image acquiring unit.

Abstract: Embodiments of the present application provide a data storage method, data acquisition method and device thereof. The method includes allocating an N-dimensional first parameter vector for N pieces of to-be-stored data; performing N-dimensional permutation on the first parameter vector, to obtain N second parameter vectors each having N dimensions; constructing a neural network model that maps the current second parameter vectors to expected data samples of the N pieces of to-be-stored data; adjusting model parameters of the neural network model and/or the first parameter vector until expected data samples of the N pieces of to-be-stored data regress to the N pieces of to-be-stored data, the expected data samples being obtained from the current second parameter vectors based on the trained neural network model; storing the current first parameter vector.

Abstract: An apparatus for acquiring an image includes an image sensor, a light compensator, and a light filter assembly. The image sensor is on a light output side of the light filter assembly, and is configured to generate and output a first image signal and a second image signal by a plurality of exposures. The first image signal is generated according to a first preset exposure and the second image signal is generated according to a second preset exposure, and they are two of the plurality of exposures. The light compensator includes a first light compensation apparatus, the first light compensation apparatus is configured to perform near-infrared light compensation, and the near-infrared light compensation is performed in at least part of an exposure period of the first preset exposure but is not performed in an exposure period of the second preset exposure.

Abstract: Embodiments of methods and apparatuses for object detection and of computer devices are disclosed.

Abstract: A surveillance method, apparatus, system, electronic device and computer-readable storage medium are provided. In the surveillance method, a non-visible light image and a target image are obtained, wherein the target image is generated from a visible light signal captured during a capture period of the non-visible light image. The method then detects whether an object is present in the non-visible light image. When an object is detected in the non-visible light image, a second location area of the object in a visible light image is determined according to a first location area of the object in the non-visible light image, such that surveillance of the object is implemented based on the visible light image, wherein the visible light image is an image determined based on the target image.

Abstract: Embodiments of the present application provide a method, apparatus, and electronic device for eliminating distortion of a distorted image. Side information components of a distorted image are generated. The distorted image is resulted from image processing on an original image. Side information components represent distortion features of the distorted image with respect to original image. Distorted image color components of the distorted image and the side information components are input into a pre-established convolutional neural network model for convolution filtering to obtain distortion-eliminated image color components. The convolutional neural network model is obtained through training based on a preset training set. The training set comprises original sample images, distorted image color components of multiple distorted images corresponding to each of the original sample images, and side information components of each of the distorted images.

Abstract: A depth measuring method and system applicable to a first binocular camera having a zoom lens is provided.

Abstract: Embodiments of the present application provide an anti-counterfeiting face detection method, device and multi-lens camera, wherein the anti-counterfeiting face detection method comprises: acquiring a depth image, an infrared image and an RGB image by using a TOF camera and an RGB camera; analyzing the RGB image through a preset face detection algorithm to determine an RGB face region of a face in the RGB image and position information of the RGB face region; determining a depth face region of the face in the depth image and an infrared face region of the face in the infrared image based on the position information of the RGB face region; determining that the face passes the detection when the depth face region, the infrared face region and the RGB face region meet corresponding preset rules respectively.

Abstract: A deep neural network operation method and apparatus are provided. The method comprises: obtaining an input feature map of a network layer; displacing respectively, according to a preset displacement parameter, each of channels of the input feature map of the network layer along axes, to obtain a displaced feature map, wherein the preset displacement parameter comprises displacement amounts of the channel in the axes; and performing convolution operation on the displaced feature map with a 1×1 convolution kernel to obtain an output feature map of the network layer. The operation efficiency of the DNN can be improved through the above method.

Abstract: The present application discloses a method and an apparatus for displaying an image based on privacy masking, which belong to the field of video monitoring. The method includes: acquiring a to-be-masked original image provided by an image capture terminal and privacy masking area information of the original image; wherein, the privacy masking area information is obtained by identifying the original image or is obtained from the image capture terminal and indicates an area in the original image which is required to be masked; determining whether the original image is required to be privacy-masked based on whether privacy unmasking permission information is stored; if the original image is required to be privacy-masked, performing privacy-masking on the original image based on the privacy masking area information to obtain a masked image, and displaying the masked image.

Abstract: Embodiments of the present application provide a method and apparatus for determining a trajectory of a human target. The method includes: extracting a target feature of a to-be-tracked human target in a to-be-processed image as a to-be-searched target feature; searching for an acquisition attribute corresponding to the to-be-searched target feature based on a pre-established correspondence between target features and acquisition attributes; determining a trajectory of the to-be-tracked human target based on the found acquisition attribute. It can be seen that in this solution, a face image is not required for determining a trajectory of a person. Even if the acquired face image is not clear, the accuracy of determining the trajectory of the person will not be reduced. Therefore, by the solution of the present application, the accuracy of determining the trajectory of the person is improved.

Abstract: The embodiments of the present application provide an image segmentation method, an image segmentation apparatus, and a fully convolutional network system. The method includes: acquiring a target image to be processed; acquiring image feature data of the target image; inputting the image feature data into a pre-trained target network far image segmentation to obtain an output; wherein the target network is a fully convolutional network comprising a hybrid context network structure, and the hybrid context network structure is configured to extract a plurality of reference features at a predetermined scale and fuse them into a target feature that matches a scale of a target object in a segmented image; and wherein the target network is trained with sample images containing target objects at different scales; and obtaining an image segmentation result for the target image based on the output.

Abstract: A method and apparatus for acquiring people flow information, and a method and apparatus for processing people flow information are provided.

Abstract: The embodiments of the present application provide a method and apparatus for recognizing an identity of a human target. The method includes: obtaining a to-be-recognized image; extracting a target feature of a to-be-recognized human target in the to-be-recognized image as a to-be-searched target feature; searching for a facial information item corresponding to the to-be-searched target feature based on a pre-established correspondence between target features and facial information items; and determining an identity of the to-be-recognized human target based on the found facial information item. It can be seen that, in this solution, there is no need to extract the facial features in the image, and even if the facial region is not clear or is shaded by other objects in the image, the accuracy of identity recognition is reduced. Therefore, accuracy of identity recognition is improved.

Abstract: An operation method and apparatus for a network layer in a Deep Neural Network are provided.