Abstract: A hardware environment-based data operation method, apparatus and device, and a storage medium. The method includes: determining data to be operated and target hardware, wherein the target hardware is a hardware resource that needs to perform convolution computation on the data to be operated currently; determining the maximum number of channels in which the target hardware executes parallel computation, and determining a data layout corresponding to the maximum number of channels to be an optimal data layout; and converting the data layout of the data to be operated into the optimal data layout, and performing the convolution computation on the data to be operated by using the target hardware after the conversion is completed.

Abstract: An image processing method, apparatus and device, and a readable storage medium are disclosed, including: obtaining a target image; inputting the target image into a quantized target deep neural network model for classification/detection to obtain an output result; and processing the target image according to a policy corresponding to the output result. A process of performing quantization to obtain the target deep neural network model includes: obtaining a pre-trained floating point type deep neural network model; extracting weight features of a deep neural network model; determining a quantization policy using the weight features; and quantizing the deep neural network model according to the quantization policy to obtain the target deep neural network model.

Abstract: A system for accelerating an RNN network including: a first cache, which is used for outputting Wx1 to WxN or Wh1 to WhN in parallel in N paths in a cyclic switching manner, and the degree of parallelism is k; a second cache, which is used for outputting xt or ht-1 in the cyclic switching manner; a vector multiplication circuit, which is used for, by using N groups of multiplication arrays, respectively calculating Wx1xt to WxNxt, or respectively calculating Wh1ht-1 to WhNht-1; an addition circuit, which is used for calculating Wx1xt+Wh1ht-1+b1 to WxNxt+WhNht-1+bN; an activation circuit, which is used for performing an activation operation according to an output of the addition circuit; a state updating circuit, which is used for acquiring ct-1, calculating ct and ht, updating ct-1, and sending ht to the second cache; a bias data cache; a vector cache; and a cell state cache.

Abstract: A neural network compression method, apparatus and device, and a storage medium are provided. The method includes: performing forward inference on target data by using a target parameter sharing network to obtain an output feature map of the last convolutional module; extracting a channel related feature from the output feature map; inputting the extracted channel related feature and a target constraint condition into a target meta-generative network; and predicting an optimal network architecture under the target constraint condition by using the target meta-generative network to obtain a compressed neural network model. By using the technical solution, the computation load of a performance evaluation process of a neural architecture search may be reduced, and the speed of the searching for a high-performance neural network architecture may be increased.

Abstract: A data processing method, system and device, and a readable storage medium. The method includes: marking each layer of a network model as a key layer or a non-key layer according to acquired structural information of the network model; respectively determining a quantization bit width range of the key layer and a quantization bit width range of the non-key layer according to hardware resource information that needs to be deployed; determining, in the quantization bit width range, optimal quantization bit widths of each layer of the network model; and training the network model based on the optimal quantization bit widths of each layer of the network model, so as to obtain an optimal network model, and performing data processing using the optimal network model.

Abstract: A hardware environment-based data quantization method includes: parsing a model file under a current deep learning framework to obtain intermediate computational graph data and weight data that are independent of a hardware environment; performing calculation on image data in an input data set through a process indicated by an intermediate computational graph to obtain feature map data; separately performing uniform quantization on the weight data and the feature map data of each layer according to a preset linear quantization method, and calculating a weight quantization factor and a feature map quantization factor (S103); combining the weight quantization factor and the feature map quantization factor to obtain a quantization parameter that makes hardware use shift instead of division; and finally, writing the quantization parameter and the quantized weight data to a bin file according to a hardware requirement so as to generate quantized file data (S105).

Abstract: Disclosed is an image processing method. In the invention, weighted calculation is carried out on a first quantification threshold value obtained with a saturated mapping method and a second quantification threshold value obtained with an unsaturated mapping method, that is, two quantification threshold values are fused, and an obtained optimal quantification threshold value can be suitable for most activation output layers, and therefore, effective information of the activation output layers can be more effectively reserved, the optimal quantification threshold value is used for subsequent image processing work, and the precision of reasoning and calculation of a quantized deep neural network on a low-bit-width hardware platform is improved. An image processing apparatus and device are disclosed, which have the same beneficial effects as the image processing method.

Abstract: A system for accelerating an RNN network including: a first cache, which is used for outputting Wx1 to WxN or Wh1 to WhN in parallel in N paths in a cyclic switching manner, and the degree of parallelism is k; a second cache, which is used for outputting xt or ht-1 in the cyclic switching manner; a vector multiplication circuit, which is used for, by using N groups of multiplication arrays, respectively calculating Wx1xt to WxNxt, or respectively calculating Wh1ht-1 to WhNht-1; an addition circuit, which is used for calculating Wx1xt+Wh1ht-1+b1 to WxNxt+WhNht-1+bN; an activation circuit, which is used for performing an activation operation according to an output of the addition circuit; a state updating circuit, which is used for acquiring ct-1, calculating ct and ht, updating ct-1, and sending ht to the second cache; a bias data cache; a vector cache; and a cell state cache.

Abstract: A hardware environment-based data quantization method includes: parsing a model file under a current deep learning framework to obtain intermediate computational graph data and weight data that are independent of a hardware environment; performing calculation on image data in an input data set through a process indicated by an intermediate computational graph to obtain feature map data; separately performing uniform quantization on the weight data and the feature map data of each layer according to a preset linear quantization method, and calculating a weight quantization factor and a feature map quantization factor (S103); combining the weight quantization factor and the feature map quantization factor to obtain a quantization parameter that makes hardware use shift instead of division; and finally, writing the quantization parameter and the quantized weight data to a bin file according to a hardware requirement so as to generate quantized file data (S105).

Abstract: Image processing method and device are disclosed. The image processing method is inclusive of a step of obtaining a target image; a step of performing motion estimation on the target image so as to attain motion paths; a step of determining, on the basis of the motion paths, whether there exists a fluctuation meeting a predetermined requirement in the target image; and a step of taking, if there exists a fluctuation meeting the predetermined requirement in the target image, the target image as a reference image. The reference image is utilized to carry out inter frame motion estimation with respect to the respective images acquired.

Abstract: Panoramic video processing method and device are disclosed. The method includes steps of receiving a panoramic video waiting for processing; conducting inter frame motion estimation in regard to any two neighboring 2D panoramic images in the panoramic video so as to obtain an inter frame motion estimation result between the same two neighboring 2D panoramic images; acquiring, based on the respective inter frame motion estimation results, a historical motion trajectory of the panoramic video, and carrying out smoothing with respect to the historical motion trajectory; calculating, based on the historical motion trajectory after smoothing, a rotation matrix for motion compensation pertaining to each 2D panoramic image in the panoramic video; and performing reconstruction on each 2D panoramic image in the panoramic video by means of the corresponding rotation matrix for motion compensation so as to attain a panoramic video after image stabilization, and outputting the panoramic video after image stabilization.

Abstract: Panoramic video processing method and device are disclosed. The method includes steps of receiving a panoramic video waiting for processing; conducting inter frame motion estimation in regard to any two neighboring 2D panoramic images in the panoramic video so as to obtain an inter frame motion estimation result between the same two neighboring 2D panoramic images; acquiring, based on the respective inter frame motion estimation results, a historical motion trajectory of the panoramic video, and carrying out smoothing with respect to the historical motion trajectory; calculating, based on the historical motion trajectory after smoothing, a rotation matrix for motion compensation pertaining to each 2D panoramic image in the panoramic video; and performing reconstruction on each 2D panoramic image in the panoramic video by means of the corresponding rotation matrix for motion compensation so as to attain a panoramic video after image stabilization, and outputting the panoramic video after image stabilization.

Abstract: Image processing method and device are disclosed. The image processing method is inclusive of a step of obtaining a target image; a step of performing motion estimation on the target image so as to attain motion paths; a step of determining, on the basis of the motion paths, whether there exists a fluctuation meeting a predetermined requirement in the target image; and a step of taking, if there exists a fluctuation meeting the predetermined requirement in the target image, the target image as a reference image. The reference image is utilized to carry out inter frame motion estimation with respect to the respective images acquired.

Abstract: A moving object detection method includes acquiring two depth image frames including depth information, which are obtained by continuously taking images of a moving object, the two depth image frames including a present depth image frame and at least one past depth image frame; dividing each of the two depth image frames into a plurality of blocks; calculating differences between numbers of pixels positioned in respective different depth areas in each of the plurality of blocks in the present depth image frame, and numbers of pixels positioned in respective different depth areas in each of the corresponding plurality of blocks in each of the at least one past depth image frame, which correspond to the plurality of blocks in the present depth image frame; and detecting a moving block in the present depth image frame based on the calculated difference and constituting the detected moving object.

Abstract: Disclosed is an object tracking method comprising steps of obtaining a grayscale image sequence and a depth image sequence; predicting a candidate target region of a tracking object in a current image; determining a depth weight image of the candidate target region; acquiring a target template which is expressed by a probability density distribution of features; expressing, by a probability density distribution of features, a candidate target whose position moves in the candidate target region; calculating, based on a probability density expression of the target template and a probability density expression of the candidate target, a similarity between the target template and the candidate target, so as to get a confidence level image of the candidate target region; and locating, based on the confidence level image of the candidate target region, a position of the tracking object in the current image.

Abstract: Disclosed is a method comprising steps of conducting image preprocessing so as to respectively obtain a candidate object region including a foreground image in a depth map and a color image; determining whether it is necessary to conduct a region growing process with respect to the candidate object region of the depth map; in the former case, conducting the region growing process with respect to the candidate object region of the depth map; and conducting, after the region growing process is conducted with respect to the candidate object region of the depth map, a depth value restoration process with respect to a candidate region.

Abstract: A method and an apparatus for tracking an object are disclosed. The method includes determining a first position by a first tracking template and determining a second position by a second tracking template, the first and second tracking templates being formed based on first and second feature sets, respectively, the first feature set being different from the second feature set, and the first feature set and the second feature set including one or more features; and determining a final position based on the first position and the second position, wherein the first tracking template is updated for each of a predetermined number of frames, the second tracking template is updated based on a predetermined rule, the second tracking template and the first tracking template are independently updated, and the update frequency of the second tracking template is lower than the first tracking template.

Abstract: Disclosed is a method comprising steps of conducting image preprocessing so as to respectively obtain a candidate object region including a foreground image in a depth map and a color image; determining whether it is necessary to conduct a region growing process with respect to the candidate object region of the depth map; in the former case, conducting the region growing process with respect to the candidate object region of the depth map; and conducting, after the region growing process is conducted with respect to the candidate object region of the depth map, a depth value restoration process with respect to a candidate region.

Abstract: Disclosed is an object tracking method comprising steps of obtaining a grayscale image sequence and a depth image sequence; predicting a candidate target region of a tracking object in a current image; determining a depth weight image of the candidate target region; acquiring a target template which is expressed by a probability density distribution of features; expressing, by a probability density distribution of features, a candidate target whose position moves in the candidate target region; calculating, based on a probability density expression of the target template and a probability density expression of the candidate target, a similarity between the target template and the candidate target, so as to get a confidence level image of the candidate target region; and locating, based on the confidence level image of the candidate target region, a position of the tracking object in the current image.

Abstract: A method and an apparatus for tracking an object are disclosed. The method includes determining a first position by a first tracking template and determining a second position by a second tracking template, the first and second tracking templates being formed based on first and second feature sets, respectively, the first feature set being different from the second feature set, and the first feature set and the second feature set including one or more features; and determining a final position based on the first position and the second position, wherein the first tracking template is updated for each of a predetermined number of frames, the second tracking template is updated based on a predetermined rule, the second tracking template and the first tracking template are independently updated, and the update frequency of the second tracking template is lower than the first tracking template.