Computing power implementation relation reasoning method and device based on intelligent computing center

Abstract

The invention provides a computing power implementation relation reasoning method and device based on an intelligent computing center, and relates to the technical field of computing power, and the method comprises the steps: S1, obtaining a to-be-evaluated image; s2, reasoning the to-be-evaluated image by using a relation reasoning model running in the intelligent computing center to obtain relation information between objects in the to-be-evaluated image; the method for reasoning the to-be-evaluated image by the relational reasoning model comprises the following steps: identifying an object in the to-be-evaluated image; extracting target features of an object in the to-be-evaluated image; establishing a spatial relationship and a semantic relationship between the objects according to the target features; and reasoning the spatial relationship and the semantic relationship between the objects to obtain relationship information between the objects in the to-be-evaluated image. According to the invention, through intelligent identification and reasoning of the spatial layout of the objects, the accuracy and efficiency of reasoning of the spatial relationship between the objects in the image are improved, and quick-response image evaluation in a scene with a high real-time requirement is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent computing centers and computing power infrastructure of intelligent computing centers, and particularly relates to a method and device for reasoning about computing power implementation relationships based on an intelligent computing center. Background Art

[0002] With the rapid development of artificial intelligence technology, "intelligent computing centers" and "intelligent computing centers" have emerged as the times require.

[0003] An "intelligent computing center" refers to a facility that provides the required computing power, data, and algorithms for artificial intelligence applications (such as scenarios like artificial intelligence deep learning model development, model training, and model inference) by using large-scale heterogeneous computing power resources, including general computing power (CPU) and intelligent computing power (GPU, FPGA, ASIC, etc.). The intelligent computing center covers facilities, hardware, and software, and can provide full-stack capabilities from underlying computing power to top-level application enabling.

[0004] The "intelligent computing center" includes, but is not limited to, the "intelligent computing center".

[0005] An "intelligent computing center", that is, an artificial intelligence computing center, is a type of computing power infrastructure that provides computing power services, data services, and algorithm services required for artificial intelligence applications based on artificial intelligence theory and using an artificial intelligence computing architecture.

[0006] "Computing power" is the ability of a computer device or a computing / data center to process information, the ability of computer hardware and software to cooperate to jointly execute a certain computing requirement, the computing ability to achieve the output of a target result by processing information data, and a new type of productive force integrating information computing power, network carrying capacity, and data storage capacity, which mainly provides services to society through computing power infrastructure.

[0007] With the continuous development of generative artificial intelligence technology, applications of image generation and analysis have gradually penetrated into multiple fields such as creative design, autonomous driving, and intelligent monitoring. In these applications, how to accurately evaluate the spatial relationships and interactions between objects in the generated images has become a key challenge.

[0008] Traditional image evaluation methods do not utilize intelligent computing centers and computing power, and usually focus on the attributes of a single object (such as color, shape, etc.), while ignoring the relationships between objects in the image. For example, when describing an image of "an apple is to the left of an orange", existing methods cannot accurately reason about the relative positions, spatial relationships, and context backgrounds of the objects. Due to the lack of an efficient relationship reasoning method, traditional technologies often struggle to meet the requirements of real-time performance and high accuracy when dealing with complex image scenarios. Summary of the Invention

[0009] The present invention provides a method and device for realizing relationship reasoning based on the computing power of an intelligent computing center to solve the problem of the lack of an efficient relationship reasoning method in existing image evaluation.

[0010] To solve the above technical problems, the present invention is implemented as follows:

[0011] In a first aspect, the present invention provides a method for realizing relationship reasoning based on the computing power of an intelligent computing center, including:

[0012] Step S1: Obtain an image to be evaluated;

[0013] Step S2: Use a relationship reasoning model running on the intelligent computing center to reason about the image to be evaluated to obtain relationship information between objects in the image to be evaluated; wherein, the relationship reasoning model reasoning about the image to be evaluated includes: identifying objects in the image to be evaluated; extracting target features of the objects in the image to be evaluated; establishing spatial relationships and semantic relationships between the objects according to the target features; reasoning about the spatial relationships and semantic relationships between the objects to obtain relationship information between objects in the image to be evaluated.

[0014] Optionally, step S1 includes:

[0015] Step S11: Obtain an original image to be evaluated;

[0016] Step S12: Perform preprocessing on the original image to be evaluated to obtain the image to be evaluated, and the preprocessing includes at least one of the following: resizing, normalization processing, data feature enhancement, and denoising.

[0017] Optionally, the relationship information includes spatial relationships and interaction relationships, wherein the spatial relationships include at least one of the following: position relationship, inclusion relationship, and contact relationship; the interaction relationships include at least one of the following: occlusion relationship, distance relationship, direction relationship, causal relationship, and behavior relationship.

[0018] Optionally, before step S2, it further includes:

[0019] Step S0: Train a relationship reasoning model to be trained;

[0020] Step S0 includes:

[0021] Step S01: Obtain training images and true relationship information between objects in the training images;

[0022] Step S02: Use the relationship reasoning model to be trained to reason about the training images to obtain predicted relationship information between objects in the training images;

[0023] Step S03: Optimize the relationship inference model to be trained according to the predicted relationship information between objects in the training image and the true relationship information between objects in the training image, so as to obtain a trained relationship inference model.

[0024] Optionally, after step S2, the following is further included:

[0025] Step S3: Visually output the relationship information between objects in the image to be evaluated in the form of image annotation or text.

[0026] Optionally, after step S2, the following is further included:

[0027] Step S4: Perform similarity matching on the relationship information between objects in the image to be evaluated and the relationship information stored in advance when generating the image to be evaluated, so as to obtain an image quality score.

[0028] In a second aspect, the present invention provides a relationship inference device based on the computing power of an intelligent computing center, including:

[0029] An acquisition module, configured to acquire an image to be evaluated;

[0030] A processing module, configured to perform inference on the image to be evaluated by using a relationship inference model running on the intelligent computing center, so as to obtain the relationship information between objects in the image to be evaluated; wherein, the relationship inference model performing inference on the image to be evaluated includes: identifying objects in the image to be evaluated; extracting target features of the objects in the image to be evaluated; establishing spatial relationships and semantic relationships between the objects according to the target features; and performing inference on the spatial relationships and semantic relationships between the objects to obtain the relationship information between objects in the image to be evaluated.

[0031] Optionally, the acquisition module includes:

[0032] A first acquisition sub-module, configured to acquire an original image to be evaluated;

[0033] A preprocessing sub-module, configured to preprocess the original image to be evaluated to obtain the image to be evaluated, and the preprocessing includes at least one of the following: adjusting the size, normalization processing, data feature enhancement, and denoising.

[0034] Optionally, the relationship information includes spatial relationships and interaction relationships, wherein the spatial relationships include at least one of the following: position relationship, inclusion relationship, and contact relationship; the interaction relationships include at least one of the following: occlusion relationship, distance relationship, direction relationship, causal relationship, and behavior relationship.

[0035] Optionally, it further includes:

[0036] A model training module for training a relationship reasoning model to be trained.

[0037] The model training module includes:

[0038] A first acquisition sub-module for acquiring the true relationship information between the training image and the objects in the training image.

[0039] An inference sub-module for inferring the training image using the relationship reasoning model to be trained, and obtaining the predicted relationship information between the objects in the training image.

[0040] An optimization sub-module for optimizing the relationship reasoning model to be trained according to the predicted relationship information between the objects in the training image and the true relationship information between the objects in the training image, and obtaining the trained relationship reasoning model.

[0041] Optionally, it further includes:

[0042] A visualization output module for visually outputting the relationship information between the objects in the image to be evaluated in the form of image annotation or text.

[0043] Optionally, it further includes:

[0044] A scoring module for performing similarity matching between the relationship information between the objects in the image to be evaluated and the relationship information stored in advance when generating the image to be evaluated, and obtaining an image quality score.

[0045] In a third aspect, the present invention provides an electronic device, including a processor, a memory, and a program or instruction stored on the memory and executable on the processor. When the program or instruction is executed by the processor, it implements the steps in the relationship reasoning method based on the computing power of the intelligent computing center as described in any item of the first aspect.

[0046] In a fourth aspect, the present invention provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, it implements the steps in the relationship reasoning method based on the computing power of the intelligent computing center as described in any item of the first aspect.

[0047] In a fifth aspect, the present invention provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, they implement the steps in the relationship reasoning method based on the computing power of the intelligent computing center as described in any item of the first aspect.

[0048] In the present invention, an image to be evaluated is obtained; a relationship reasoning model running in an intelligent computing center is used to reason about the image to be evaluated, and relationship information between objects in the image to be evaluated is obtained; wherein, the relationship reasoning model reasoning about the image to be evaluated includes: identifying objects in the image to be evaluated; extracting target features of the objects in the image to be evaluated; establishing spatial relationships and semantic relationships between the objects according to the target features; reasoning about the spatial relationships and semantic relationships between the objects to obtain the relationship information between the objects in the image to be evaluated. In this way, through the powerful computing power provided by the intelligent computing center, the relationship reasoning model is used to intelligently identify and reason about the spatial layout of objects, improving the accuracy and efficiency of reasoning about the spatial relationships between objects in the image, realizing image evaluation with fast response in scenarios with high real-time requirements, and solving the problem of the lack of an efficient relationship reasoning method in existing image evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0050] Figure 1 is a flowchart of a relationship reasoning method implemented based on the computing power of an intelligent computing center provided by the present invention;

[0051] Figure 2 is a schematic structural diagram of a relationship reasoning device implemented based on the computing power of an intelligent computing center provided by the present invention;

[0052] Figure 3 is a schematic structural diagram of an electronic device provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0054] The "computing power" described in the present invention is the ability of a computer device or a computing / data center to process information, the ability of computer hardware and software to cooperate to execute a certain computing requirement together, the computing ability to realize the output of a target result by processing information data, a new type of productive force integrating information computing power, network carrying capacity, and data storage capacity, and mainly provides services to society through computing power infrastructure.

[0055] The "Computational Power (CP)" described in the present invention is a capability of a data center server to process data and output results, which is a comprehensive indicator for measuring the computing power of a data center and includes general computing power, supercomputing power, and intelligent computing power. The commonly used measurement unit is the number of floating-point operations per second (FLOPS, 1 EFLOPS = 10^18 FLOPS). The larger the value, the stronger the comprehensive computing power. It is estimated that 1 EFLOPS is approximately the computing power output of 5 Tianhe-2A or 500,000 mainstream server CPUs or 2 million mainstream laptops. The calculation formula is: CP = CP_general + CP_intelligent + CP_super.

[0056] The "Network Power (NP)" described in the present invention is the manifestation of the data transmission capability of computing power facilities, which is a comprehensive capability including network architecture, network bandwidth, transmission delay, intelligent management and scheduling, etc., and involves network transmission within and between data centers. It is a comprehensive indicator for measuring network transmission scheduling capabilities.

[0057] The "Storage Power (SP)" described in the present invention is the comprehensive capability of a data center in terms of data storage capacity, performance, security and reliability, and green and low-carbon aspects. It is a comprehensive indicator for measuring the data storage capacity of a data center and includes external storage devices such as storage arrays and server-internal storage devices. The commonly used measurement unit for storage capacity is exabyte (EB, 1 EB = 2^60 bytes), and the commonly used measurement unit for performance is the number of read / write operations per second per unit capacity (IOPS / TB, Input / Output Operations Per Second / TB). The disaster recovery ratio is an important manifestation of security and reliability.

[0058] The "computing power infrastructure" described in the present invention refers to a new type of information infrastructure that integrates information computing power, network carrying power, and data storage power, and can realize the centralized computing, storage, transmission, and application of information, presenting characteristics such as multi-element ubiquitous, intelligent and agile, secure and reliable, and green and low-carbon.

[0059] The "new type of information infrastructure" described in the present invention mainly includes network infrastructures such as 5G networks, fiber broadband networks, backbone networks, international communication networks, and satellite Internet, computing power infrastructures such as data centers, general computing power centers, intelligent computing centers, and supercomputing centers, and new technology facilities such as artificial intelligence, blockchain, and quantum computing. With the emergence and popularization of new general technologies, the form of the new type of information infrastructure will be more diverse.

[0060] The "computing power" described in the present invention includes: general computing power, intelligent computing power, and supercomputing power.

[0061] The "general computing power" described in the present invention refers to the computing power provided by servers based on CPU (Central Processing Unit) chips, which is used to support basic general computing such as cloud computing and edge computing.

[0062] The "intelligent computing power" described in the present invention refers to the computing platforms deployed on a large scale for various artificial intelligence innovation applications based on dedicated chips such as GPU (Graphics Processing Unit), FPGA (Field Programmable Gate Array), and ASIC (Application Specific Integrated Circuit), such as natural language processing, machine vision, and so on.

[0063] The "super computing power" described in the present invention mainly refers to the computing power provided by high-performance computing clusters such as supercomputers. It utilizes the centralized computing resources of multiple computer systems working in parallel and processes extremely complex or data-intensive problems through a dedicated operating system. It is mainly used for computing in cutting-edge scientific fields, such as planetary simulation, drug molecule design, gene analysis, etc.

[0064] The "intelligent computing center" described in the present invention refers to a facility that provides the required computing power, data, and algorithms for artificial intelligence applications (such as scenarios of artificial intelligence deep learning model development, model training, and model inference) by using large-scale heterogeneous computing power resources, including general computing power (CPU) and intelligent computing power (GPU, FPGA, ASIC, etc.). The intelligent computing center covers facilities, hardware, and software, and can provide full-stack capabilities from underlying computing power to top-level application enabling.

[0065] The "intelligent computing center" described in the present invention includes, but is not limited to, the "intelligent computing center".

[0066] The "intelligent computing center" described in the present invention, that is, the artificial intelligence computing center, is a type of computing power infrastructure that provides computing power services, data services, and algorithm services required for artificial intelligence applications based on artificial intelligence theory and using an artificial intelligence computing architecture.

[0067] The "computing power center" described in the present invention refers to a facility mainly composed of infrastructure such as wind, fire, water, and electricity and IT software and hardware devices, which has computing power, transportation power, and storage power, including general data centers, intelligent computing centers, supercomputing centers, etc.

[0068] The "supercomputing center" described in the present invention refers to: namely, a supercomputing data center, which is a data center based on supercomputers or large-scale computing clusters and can provide functions such as large-scale computing, storage, and network services, and is widely used in application scenarios such as aerospace, national defense, oil exploration, climate modeling, and genome sequencing.

[0069] The "computing power resources" described in the present invention refers to: the technologies and facilities with information computing, transmission, storage, and application capabilities required for the development of the digital society, including but not limited to computing resources such as CPUs and GPUs, network resources such as switches and routers, storage resources such as storage arrays and distributed storage, security resources such as firewalls and intrusion detection systems, and support and guarantee resources such as wind, fire, water, and electricity.

[0070] The "relationship" described in the present invention refers to: in image generation and analysis, the mutual connection and interaction patterns in aspects such as space, time, and function between different objects or elements in the image. Specifically, a relationship is not just an individual attribute of an object, but rather the relative position, direction, action, etc. between objects. For example, objects in an image may have a positional relationship in space, or a causal relationship in time, or even a certain dependency relationship in semantics.

[0071] The "relationship reasoning" described in the present invention refers to: the process of analyzing and understanding the relationships between multiple objects in an image. This reasoning comprehensively interprets the spatial information, semantic information, and context of the image to reveal the relative positions, interaction methods, and logical connections between objects. The goal of relationship reasoning is to abstract high-level semantic information from image data, enabling the machine to understand complex scenes in the image in a way similar to humans.

[0072] Please refer to Figure 1 , the present invention provides a method for realizing relationship reasoning based on the computing power of an intelligent computing center, including:

[0073] Step S1: Obtain the image to be evaluated;

[0074] In an embodiment of the present invention, optionally, the step S1 includes:

[0075] Step S11: Obtain the original image to be evaluated;

[0076] Step S12: Perform preprocessing on the original image to be evaluated to obtain the image to be evaluated, and the preprocessing includes at least one of the following: adjusting the size, normalizing, enhancing data features, and denoising.

[0077] In an embodiment of the present invention, performing preprocessing on the original image to be evaluated provides clearer and more standardized image data for subsequent analysis and processing, improves the quality and consistency of the image data, and thus enhances the effect of subsequent processing and analysis.

[0078] In the embodiments of the present invention, the image to be evaluated may be an image directly uploaded or an image generated by artificial intelligence. The image to be evaluated includes different objects or elements, and the spatial, temporal, functional, and other interrelationships and interaction patterns among different objects or elements in the image. Specifically, the relationship is not only an individual attribute of an object, but also the relative position, direction, action, and other relationships between objects. For example, the objects in the image may have a positional relationship in space, a causal relationship in time, or even a certain dependency relationship in semantics.

[0079] Step S2: Use the relationship reasoning model running in the intelligent computing center to reason about the image to be evaluated, and obtain the relationship information between the objects in the image to be evaluated; wherein, the relationship reasoning model reasoning about the image to be evaluated includes: identifying the objects in the image to be evaluated; extracting the target features of the objects in the image to be evaluated; establishing the spatial relationship and semantic relationship between the objects according to the target features; reasoning about the spatial relationship and semantic relationship between the objects to obtain the relationship information between the objects in the image to be evaluated.

[0080] In the embodiments of the present invention, optionally, before the step S2, it further includes:

[0081] Step S0: Perform model training on the relationship reasoning model to be trained;

[0082] The step S0 includes:

[0083] Step S01: Obtain the true relationship information between the training image and the objects in the training image;

[0084] Step S02: Use the relationship reasoning model to be trained to reason about the training image, and obtain the predicted relationship information between the objects in the training image;

[0085] Step S03: Optimize the relationship reasoning model to be trained according to the predicted relationship information between the objects in the training image and the true relationship information between the objects in the training image, and obtain the trained relationship reasoning model.

[0086] In the embodiments of the present invention, a training image is used to train the relationship reasoning model, and the parameters of the model are adjusted to minimize the loss function. Among them, common optimization algorithms can be used but are not limited to gradient descent or Adaptive Moment Estimation (Adam), etc. And the performance of the model can be monitored, and the model can be updated regularly to cope with changes in data distribution or the introduction of new data. Among them, the relationship reasoning model can, but is not limited to, use a pre-trained visual model to extract the target features of the objects in the image to be evaluated, and combine object detection and relationship reasoning to identify the specific relationships between the objects in the picture.

[0087] In the embodiments of the present invention, optionally, the relationship information includes spatial relationship and interaction relationship. Among them, the spatial relationship includes at least one of the following: position relationship, inclusion relationship, and contact relationship; the interaction relationship includes at least one of the following: occlusion relationship, distance relationship, direction relationship, causal relationship, and behavior relationship.

[0088] In the embodiments of the present invention, a relationship reasoning model running in the intelligent computing center is adopted. Through the powerful computing power provided by the intelligent computing center, it supports efficient reasoning in complex scenarios and can flexibly adapt to different application scenarios, such as object relationship recognition in virtual reality and autonomous driving scenarios, intelligent monitoring, etc. And a relationship reasoning model is adopted to perform inferences such as position inference, spatial inference, and context inference to meet different image evaluation requirements.

[0089] In the embodiments of the present invention, optionally, after step S2, it further includes:

[0090] Step S3: Visualize and output the relationship information between the objects in the image to be evaluated in the form of image annotation or text.

[0091] In the embodiments of the present invention, the inference result, that is, the relationship information between the objects in the image to be evaluated, is output in a visual way, including an image annotating the objects and their relationships, or describing the relationships between the objects in text form, improving the transparency and usability of the model, and being able to more clearly display the evaluation result of the image to be evaluated, facilitating further evaluation by subsequent manual or system, so as to promote better business results.

[0092] In the embodiments of the present invention, optionally, after step S2, it further includes:

[0093] Step S4: Perform similarity matching between the relationship information between the objects in the image to be evaluated and the relationship information stored in advance when generating the image to be evaluated to obtain an image quality score.

[0094] In the embodiments of the present invention, the relationship reasoning model may, but is not limited to, use a reward model to perform similarity matching between the relationship information between the objects in the image to be evaluated and the relationship information stored in advance when generating the image to be evaluated, so as to obtain an image quality score;

[0095] In some embodiments, taking the image to be evaluated as coming from an Artificial Intelligence Generated Content (AIGC) generation model as an example, the reward model optimizes the framework output by the generation model through a reward signal. That is, in the evaluation task of the image to be evaluated generated by AIGC, the goal of the reward model is to automatically measure the quality of the image to be evaluated and provide meaningful feedback to guide the generation model (such as a diffusion model) to generate higher-quality pictures. Specifically, the quality evaluation here is mainly based on the reasoning ability of the relationships between the objects or elements in the picture. In other words, the task of the reward model is to analyze whether the relationship information of the objects in the generated picture meets the expectations and score or reward based on this.

[0096] Specifically, by calculating the similarity between the relationship information A between the objects in the image to be evaluated and the relationship information B stored in advance when generating the image to be evaluated: Rattr = Sim(A, B); where Sim(·) can use cosine similarity, Euclidean distance, or other feature comparison methods;

[0097] Then, the multi-modal consistency reward is obtained using the semantic similarity score between the text and the picture: Rtext-img = Sim(CLIPimg(image), CLIPtext(input)); where CLIPimg(image) is the first feature vector obtained by inputting the image image into the image encoder of the multi-modal pre-training model (Contrastive Language-Image Pre-training, CLIP), representing the semantic information of the picture; CLIPtext(input) is the second feature vector obtained by inputting the text input into the text encoder of CLIP, representing the semantic information of the text.

[0098] Traditional image quality evaluation metrics can be introduced, such as Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity (SSIM), or a dedicated network can be trained to predict the subjective quality score of the picture, so as to integrate various evaluation metrics to calculate the final reward score. In the subsequent reinforcement learning stage, the loss function for guiding the improvement of the generation model is obtained through the score in the reward model: LossRL = -Epolicy[R(image)], where Epolicy[R(image)] is the reward score obtained for the image generated by the model under the current policy.

[0099] In the embodiments of the present invention, by designing a reward function, the picture quality is comprehensively evaluated from dimensions such as the matching degree between relationships, multimodal consistency, and visual quality, and the reward model is trained through supervised learning. The reward model is combined with the generation model, and reinforcement learning is used to optimize the generation process to improve the quality, attribute consistency, and diversity of picture generation.

[0100] Please refer to Figure 2 , the embodiments of the present invention provide a relationship reasoning device based on the computing power of an intelligent computing center, including:

[0101] An acquisition module 21, configured to acquire an image to be evaluated;

[0102] A processing module 22, configured to perform reasoning on the image to be evaluated by using a relationship reasoning model running on the intelligent computing center to obtain relationship information between objects in the image to be evaluated; wherein, the relationship reasoning model performing reasoning on the image to be evaluated includes: identifying objects in the image to be evaluated; extracting target features of the objects in the image to be evaluated; establishing spatial relationships and semantic relationships between the objects according to the target features; and reasoning on the spatial relationships and semantic relationships between the objects to obtain relationship information between the objects in the image to be evaluated.

[0103] In the embodiments of the present invention, optionally, the acquisition module includes:

[0104] A first acquisition sub-module, configured to acquire an original image to be evaluated;

[0105] A preprocessing sub-module, configured to preprocess the original image to be evaluated to obtain the image to be evaluated, and the preprocessing includes at least one of the following: resizing, normalization processing, data feature enhancement, and denoising.

[0106] In an embodiment of the present invention, optionally, the relationship information includes a spatial relationship and an interaction relationship, where the spatial relationship includes at least one of the following: a position relationship, an inclusion relationship, and a contact relationship; the interaction relationship includes at least one of the following: an occlusion relationship, a distance relationship, a direction relationship, a causal relationship, and a behavior relationship.

[0107] In an embodiment of the present invention, optionally, it further includes:

[0108] A model training module, configured to perform model training on a relationship inference model to be trained;

[0109] The model training module includes:

[0110] A first acquisition sub-module, configured to acquire the true relationship information between the training image and the objects in the training image;

[0111] An inference sub-module, configured to perform inference on the training image by using the relationship inference model to be trained, and obtain the relationship information between the objects in the predicted training image;

[0112] An optimization sub-module, configured to optimize the relationship inference model to be trained according to the relationship information between the objects in the predicted training image and the true relationship information between the objects in the training image, and obtain a trained relationship inference model.

[0113] In an embodiment of the present invention, optionally, it further includes:

[0114] A visualization output module, configured to visually output the relationship information between the objects in the image to be evaluated in the form of image annotation or text.

[0115] In an embodiment of the present invention, optionally, it further includes:

[0116] A scoring module, configured to perform similarity matching on the relationship information between the objects in the image to be evaluated and the relationship information stored in advance when generating the image to be evaluated, and obtain an image quality score.

[0117] The relationship inference based on the computing power of the intelligent computing center provided by the embodiment of the present invention can implement Figure 1 each process implemented by the method embodiment and achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0118] An embodiment of the present invention provides an electronic device 30, as shown in Figure 3 shown, Figure 3The following is a block diagram of the principle of the electronic device 30 according to an embodiment of the present invention, which includes a processor 31, a memory 32, and a program or instruction stored on the memory 32 and executable on the processor 31. When the program or instruction is executed by the processor, the steps in any of the relationship reasoning methods based on the computing power of the intelligent computing center according to the present invention are implemented.

[0119] An embodiment of the present invention provides a readable storage medium with a program or instruction stored thereon. When the program or instruction is executed by a processor, the various processes of the embodiments of the relationship reasoning method based on the computing power of the intelligent computing center as described above are implemented, and the same technical effects can be achieved. To avoid repetition, details are not described herein again.

[0120] An embodiment of the present invention also provides a computer program product, including a computer instruction, which, when executed by a processor, implements the various processes of the method embodiment as described above Figure 1 and can achieve the same technical effects. To avoid repetition, details are not described herein again.

[0121] Computer-readable media include both permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory, or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD), or other optical storage, magnetic cassette tapes, magnetic disk storage, or other magnetic storage devices, or any other non-transmission media that can be used to store information accessible by a computing device. As defined in the present invention, computer-readable media do not include transitory media such as modulated data signals and carrier waves.

[0122] It should be noted that in the technical solution of the present invention, the collection, acquisition, update, analysis, processing, use, transmission, storage, etc. of the user's personal information comply with the provisions of relevant laws and regulations, are used for legal purposes, and do not violate public order and good customs. Necessary measures are taken for the user's personal information to prevent illegal access to the user's personal information data and to maintain the security of the user's personal information and network security.

[0123] It should be noted that in the present invention, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the phrase "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising such element.

[0124] The serial numbers of the above embodiments of the present invention are only for description and do not represent the superiority or inferiority of the embodiments.

[0125] Through the description of the above embodiments, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present application, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions for causing a service classification device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in various embodiments of the present application.

[0126] The above are only the preferred embodiments of the present application. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A method for realizing relational reasoning based on the computing power of an intelligent computing center, characterized in that: include: Step S1: Obtain an image to be evaluated; Step S2: using the relational reasoning model running in the intelligent computing center to reason on the image to be evaluated, and obtain the relationship information between the objects in the image to be evaluated; wherein the relational reasoning model reasoning on the image to be evaluated includes: identifying the objects in the image to be evaluated; extracting the target features of the objects in the image to be evaluated; establishing the spatial relationship and semantic relationship between the objects according to the target features; reasoning the spatial relationship and semantic relationship between the objects, and obtain the relationship information between the objects in the image to be evaluated.

2. The method for realizing relational reasoning based on the computing power of an intelligent computing center according to claim 1, characterized in that: The step S1 comprises: Step S11: obtaining the original image to be evaluated; Step S12: preprocessing the original image to be evaluated to obtain the image to be evaluated, wherein the preprocessing includes at least one of the following: resizing, normalization, data feature enhancement, and denoising.

3. The method for realizing relational reasoning based on the computing power of an intelligent computing center according to claim 1, characterized in that: The relationship information includes spatial relationships and interactive relationships, wherein the spatial relationship includes at least one of the following: position relationship, containment relationship and contact relationship; the interactive relationship includes at least one of the following: occlusion relationship, distance relationship, direction relationship, cause-effect relationship and behavior relationship.

4. The method for realizing relational reasoning based on the computing power of an intelligent computing center according to claim 1, characterized in that: Before step S2, the method further includes: Step S0: Perform model training on the relational reasoning model to be trained; The step S0 comprises: Step S01: obtaining real relationship information between a training image and an object in the training image; Step S02: using the relational reasoning model to be trained to reason on the training image to obtain predicted relational information between objects in the training image; Step S03: optimizing the relational reasoning model to be trained according to the predicted relational information between the objects in the training image and the actual relational information between the objects in the training image to obtain a trained relational reasoning model.

5. The method for realizing relational reasoning based on the computing power of an intelligent computing center according to claim 1, characterized in that: The step S2 further includes: Step S3: Visually output the relationship information between the objects in the image to be evaluated in the form of image annotation or text.

6. The method for realizing relational reasoning based on the computing power of an intelligent computing center according to claim 1, characterized in that: The step S2 further includes: Step S4: performing similarity matching on the relationship information between the objects in the image to be evaluated and the pre-stored relationship information stored when the image to be evaluated is generated, to obtain an image quality score.

7. A device for realizing relational reasoning based on the computing power of an intelligent computing center, characterized in that: include: An acquisition module, used for acquiring an image to be evaluated; A processing module is used to use the relational reasoning model running in the intelligent computing center to reason about the image to be evaluated, so as to obtain the relationship information between the objects in the image to be evaluated; wherein the relational reasoning model reasoning about the image to be evaluated includes: identifying the objects in the image to be evaluated; extracting the target features of the objects in the image to be evaluated; establishing the spatial relationship and semantic relationship between the objects according to the target features; and reasoning about the spatial relationship and semantic relationship between the objects to obtain the relationship information between the objects in the image to be evaluated.

8. An electronic device, characterized in that: It includes a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps in the method for realizing relational reasoning based on the computing power of an intelligent computing center as described in any one of claims 1 to 6.

9. A readable storage medium, characterized in that: The readable storage medium stores programs or instructions, and when the programs or instructions are executed by the processor, the steps in the method for realizing relational reasoning based on the computing power of an intelligent computing center as described in any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that It includes computer instructions, which, when executed by a processor, implement the steps in the method for realizing relational reasoning based on the computing power of an intelligent computing center as described in any one of claims 1 to 6.

Images