Inspur Suzhou Intelligent Technology Co., Ltd

Suzhou Metabrain Intelligent Technology Co., Ltd.

Inspur Electronic Information Industry Co., Ltd

Shandong Yingxin Computer Technologies Co., Ltd.

Yaqian Zhao

Rengang Li

Baoyu Fan

Gang Dong

Li Wang

Haiwei Liu

Wenfeng Yin

Hongbin Yang

Xiaochuan Li

Dongdong Jiang

Xin Zhang

Zhenhua Guo

Lingyan Liang

Ruyang Li

Qichun Cao

Nan Wu

Hongzhi Shi

Runze Zhang

Chen Li

Yaqiang Zhang

Xiangyi Deng

Hui Wei

Jinzhe Jiang

Xiaomin Zhu

A natural language processing method and apparatus, a device, and a readable storage medium, where the method includes: obtaining a target sentence to be processed, and determining each entity in the target sentence (S); for each entity in the target sentence, in response to the entity being present in a preset entity set, determining extended information for the entity, and adding the determined extended information after a location of the entity in the target sentence, to obtain an updated target sentence (S); and inputting the updated target sentence to a bidirectional encoder representations from transformer (BERT) model, such that the BERT model performs a natural language processing task, where in a process in which the BERT model performs the natural language processing task, an attention score between extended information of any entity in the target sentence and another entity in the target sentence is adjusted to zero (S).

Natural language processing method and apparatus, device, and readable storage medium using a three dimensional tensor to determine a relation between entities to aid in a question-answering task

A three-dimensional reconstruction method, a system, and a non-transitory computer readable storage medium are disclosed herein. The method includes: performing local pose optimization by using a target image frame to obtain a local pose error; performing neural network prediction on the target image frame to obtain an initial reconstruction error; performing three-dimensional reconstruction according to the local pose error and the initial reconstruction error to obtain an initial reconstruction model; performing global pose optimization by using historical image frames to obtain a global optimization result and a global pose error; performing neural network completion on the global optimization result to obtain a final reconstruction error; and optimizing the initial reconstruction model according to the global pose error and the final reconstruction error to obtain a final reconstruction model. The three-dimensional reconstruction method can achieve high-quality and high-precision three-dimensional reconstruction more quickly and conveniently.

Three-dimensional reconstruction method, system, and storage medium

An image recognition method includes: inputting a sample image data set into an original neural network model in advance; and for each convolution layer of the original neural network model, using a feature map of the sample image data set in the current layer as a reconstruction target, firstly obtaining an update weight of a convolution kernel by using a kernel set construction method, then calculating an input channel combination having a minimum reconstruction error, cutting a redundant input channel to obtain a compression result of the current convolution layer, and finally, splicing compression results of the convolution layers to generate an image recognition model. An image to be recognized is obtained and is inputted into the image recognition model, and an output result of the image recognition model is used as an image recognition result of the image.

Image recognition method and device, and computer-readable storage medium

The present application relates to the technical field of artificial intelligence, and discloses a visual positioning method and apparatus, a device, and a medium. The method includes: performing feature splicing on an image encoding feature and a text encoding feature; performing feature fusion on spliced encoding features to obtain a first fused encoding feature; performing noise correction on the first fused encoding feature and the text encoding feature on the basis of a preset cross-attention mechanism to obtain a corrected fused feature and a corrected text encoding feature, and performing feature fusion on the spliced encoding feature and the corrected text encoding feature to obtain a second fused encoding feature; and correcting a preset frame feature using a target encoding feature on the basis of the corrected fused feature and the second fused encoding feature to predict a regional position coordinate of a target visual object.

Visual positioning method and apparatus, device, and medium

Embodiments of the present application disclose a natural language processing method and apparatus, a device, and a readable storage medium. The method includes: obtaining a target sentence to be processed, and determining each first entity in the target sentence; for each first entity, in response to the first entity being present in a preset entity set, determining, in the preset entity set, a second entity in maximum correlation with the first entity, generating extended information based on the determined second entity, and adding the extended information after a location of the first entity in the target sentence, to obtain an updated target sentence, where the second entity is any entity in the preset entity set other than the first entity; and inputting the updated target sentence to a bidirectional encoder representations from transformer (BERT) model, such that the BERT model performs a natural language processing task.

Natural language processing method and apparatus, device, and readable storage medium

Disclosed is an image segmentation method, including: obtaining an original image set; performing feature extraction on the original image set by using a backbone network to obtain a feature map set; performing channel extraction fusion processing on the feature map set by using a channel extraction fusion model to obtain an enhanced feature map set; and segmenting the enhanced feature map set by using a preset convolutional neural network to obtain an image segmentation result. In addition, the present application also provides an image segmentation system and device, and a readable storage medium, which have the beneficial effects above.

Image segmentation method, system, and device, and readable storage medium

The present application discloses a person re-identification method, system and device, and a computer-readable storage medium. The method includes: obtaining a first type of person image without a label; making label information for the first type of person image; training a target person re-identification network on the basis of the first type of person image and the label information to obtain a first trained target person re-identification network; extracting a region of interest in the first type of person image; and training, on the basis of the first type of person image and the region of interest, the first trained target person re-identification network to obtain a second trained target person re-identification network, and performing person re-identification on the basis of the target person re-identification network.

Person re-identification method, system and device, and computer-readable storage medium

A difference description statement generation method includes: performing feature concatenation on an image encoding feature and a text encoding feature; inputting a concatenated encoding feature to a preset image-text alignment unit constructed based on a preset self-attention mechanism to obtain a concatenated alignment feature; processing, by using a preset noise monitoring unit constructed based on the preset self-attention mechanism and a preset cross-attention mechanism, the text encoding feature, and an image alignment feature and a text alignment feature that are obtained by splitting the concatenated alignment feature respectively, to extract a difference signal; and generating a difference description statement based on the difference signal by using a preset difference description generation algorithm.

Difference description statement generation method and apparatus, device and medium

An image processing method is provided. Since in the present invention, the first quantization threshold obtained by the saturated mapping method and the second quantization threshold obtained by the unsaturated mapping method are weighted, it is equivalent to fusing two quantization threshold values. The obtained optimal quantization threshold can be applied to most activations, to more effectively retain the effective information of the activations and use in subsequent image processing, thereby improving the accuracy of inference computations of the quantified deep neural network on low-bit-width hardware platforms. An image processing device and apparatus have the same beneficial effects as the above image processing method.

Image processing method, device and apparatus

A high-resolution image matching method and system are provided. The method includes performing regional fidelity down-sampling on an initial high-resolution image to obtain a multi-level low-resolution image, performing local matching on the obtained multi-level low-resolution images using a method with global probes to obtain a matching result of the low-resolution images, and performing reverse refinement on the obtained matching result of the low-resolution image using overall consistency of the image matching, to obtain the matching results of the high-resolution images at all levels until the matching results of the initial resolution images are obtained, so as to reduce the computational complexity of the image matching process and improve the accuracy of the matching result, and then the matching result of the high-resolution image is obtained through reverse refinement based on the overall consistency.

High-resolution image matching method and system

A neural network compression method whereby forward inference is performed on target data by using a target parameter sharing network to obtain an output feature map of the last convolutional module, a channel related feature is extracted from the output feature map, the extracted channel related feature and a target constraint condition are input into a target meta-generative network, and an optimal network architecture under the target constraint condition is predicted by using the target meta-generative network to obtain a compressed neural network model.

Neural network compression method, apparatus and device, and storage medium

A task scheduling method includes: when a task requirement is obtained, splitting the task requirement to obtain the plurality of subtasks having a constraint relationship; performing execution condition detection on non-candidate subtasks, determining a non-candidate subtask that satisfies an execution condition as a candidate subtask, and putting the candidate subtask into a task queue; performing state detection on a server network composed of edge servers to obtain server state information and communication information; inputting the server state information, the communication information, and queue information corresponding to the task queue into an action value evaluation model to obtain the plurality of evaluated values respectively corresponding to the plurality of scheduling actions; and determining a target scheduling action from the plurality of scheduling actions by using the evaluated values, and scheduling the candidate subtask in the task queue on the basis of the target scheduling action.

Task scheduling method and apparatus, electronic device, and readable storage medium

The present application discloses an automatic driving control method. In the method, parameters are optimally set by using a noisy and noiseless dual-strategy network, identical vehicle traffic environment state information is input into the noisy and noiseless dual-strategy network, a motion space perturbation threshold is set by using a noiseless strategy network as a comparison and a benchmark so as to adaptively adjust noise parameters, and motion noise is indirectly added by adaptively injecting noise into a strategy network parameter space, such that exploration of an environment and a motion space by a deep reinforcement learning algorithm may be effectively improved, automatic driving exploration performance and stability based on deep reinforcement learning is improved, and full consideration of influence of an environment state and driving strategies in vehicle decision-making and motion selection is ensured, thereby improving the stability and safety of an automatic vehicle.

Autonomous driving control method, apparatus and device, and readable storage medium

An image recognition method and apparatus based on a systolic array, and a medium are disclosed. The method includes: converting obtained image feature information into a one-dimensional feature vector; converting an obtained weight matrix into a one-dimensional weight vector, and allocating a corresponding weight group to each node in a trained three-dimensional systolic array model; performing multiply-accumulate of the feature vector and a weight value on the one-dimensional feature vector in parallel by using the three-dimensional systolic array model, to obtain a feature value corresponding to each node, the feature values with different values reflecting article categories contained in an image; and determining an article category contained in the image according to the feature value corresponding to each node and a pre-established corresponding relationship between the feature value and the article category.

Image recognition method and apparatus based on systolic array, and medium

The present invention provides a method and apparatus for data caching. The method comprises: output matrixes are acquired one by one, a plurality of acquired output matrixes are written alternately into two queue sets of a first cache unit according to a sequence in which the output matrixes are acquired, and the output matrixes stored line by line in a first cache unit are written into a second cache unit one by one, according to the sequence in which the output matrixes are written into the second cache unit, valid data of each output matrix of the second cache unit is determined one by one according to preset parameters, and the valid data of each output matrix is written into a third cache unit, and the valid data of the output matrixes stored in the third cache unit are configured to be sequentially written into a memory according to a sequence in which the valid data are written into the third cache unit. In the present solution, the output matrixes are cached by using cache units with the writing speed matching with the computing speed of a processor, and the output matrixes are completely written into a memory one by one according to a sequence of generation time. Therefore, the present invention may solve the problem that the computing speed of the processor does not match with the writing speed of the memory.

Growing community of inventors

Jiangsu, China

Yaqian Zhao