WO2025180090A1 - Data processing method and apparatus - Google Patents

Abstract

A data processing method, which is applied in the field of artificial intelligence. The method comprises: acquiring features of a video, wherein the features comprise a plurality of feature maps, which are distributed in a temporal dimension, and the dimensions of the feature maps comprise a channel dimension and a spatial dimension; changing the plurality of feature maps from being distributed in the temporal dimension to being distributed in the channel dimension or the spatial dimension, so as to obtain one or more first feature maps, wherein the first feature maps do not comprise the temporal dimension; and processing the first feature maps by means of a feature extraction network, so as to obtain one or more second feature maps. By means of the present application, the dimensions of features of a video are reduced, and when the dimensions of the features are reduced, a subsequently required computation amount can be greatly reduced.

Description

A data processing method and device thereof

This application claims priority to the Chinese patent application filed with the State Intellectual Property Office on March 1, 2024, with application number 202410239097.5 and application name “A data processing method and device thereof”, the entire contents of which are incorporated by reference into this application.

Technical Field

The present application relates to the field of artificial intelligence, and in particular to a data processing method and device thereof.

Background Art

Artificial Intelligence (AI) is the theory, methods, techniques, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, to perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a branch of computer science that seeks to understand the essence of intelligence and produce new intelligent machines that can respond in a manner similar to human intelligence. AI also involves studying the design principles and implementation methods of various intelligent machines, enabling them to perceive, reason, and make decisions.

Unlike image recognition, video processing tasks (such as video understanding) require more computing resources. Therefore, it is crucial to research lightweight video understanding models to reduce resource consumption. Lightweight video understanding models have many practical applications, such as autonomous driving, robotics, and industrial control. However, existing video understanding models focus more on achieving higher accuracy, resulting in relatively large model designs, while paying little attention to improving the performance of small models in end-to-end applications.

Therefore, there is an urgent need for a lightweight model design for video processing.

Summary of the Invention

In a first aspect, the present application provides a data processing method, which can be executed by a video processing system, and the method includes: the video processing system obtains features of a video, wherein the features include multiple feature maps distributed in a time dimension, and the dimensions of the feature maps include a channel dimension and a spatial dimension, that is, the features of the video are at least four-dimensional features, and too high a dimension will result in excessive computational overhead for subsequent operations; the video processing system converts the multiple feature maps from a distribution in the time dimension to a distribution in the channel dimension or the spatial dimension to obtain the one or more first feature maps; and based on the one or more first feature maps, obtains the task processing result of the video.

Among them, changing the distribution of the multiple feature maps from the time dimension to the channel dimension or the spatial dimension can be understood as: fusing the multiple feature maps in the time dimension to the channel dimension or the spatial dimension (fusion can be described as compression, for example, fusion to the channel dimension, or fusion to the spatial dimension, or fusion to the channel dimension and the spatial dimension at the same time).

Existing video understanding models have poor performance on mobile terminals, while models with better performance consume too much resources, which is not conducive to application to end-side devices. It is difficult to balance the resource consumption and performance of mobile terminal models. The idea of this application is to reduce the dimension of video features. Specifically, the features in the time dimension are compressed to the channel dimension or the spatial dimension, so that the compressed features (that is, the first feature map in the embodiment of this application) do not include the time dimension, and the amount of data in the channel dimension and the spatial dimension increases. When the feature dimension is reduced, the amount of subsequent calculations required can be greatly reduced, for example, it can be changed from 3D convolution to 2D convolution.

The first feature map does not include the time dimension.

In which, when obtaining the task processing result of the video based on the one or more first feature maps, the first feature map can be processed by a feature extraction network to obtain one or more second feature maps, and the task processing result of the video can be obtained through the task network based on the one or more second feature maps.

In a possible implementation, the task processing result is a processing result of a video understanding task, a video-based generation task, or a video enhancement task.

In a possible implementation, the first feature map and the feature have the same size in spatial dimension.

In one possible implementation, the size of the feature is (x, t, h, w), and the size of the first feature map is (x*t, h, w), where x is the number of channels, t is the number of times, h and w are the height and width in the spatial dimension, respectively, and * is the product.

In a possible implementation, the feature maps in different time dimensions among the multiple feature maps may be stacked in the channel dimension to obtain the one or more first feature maps, and the stacking order may be based on the order in the time dimension.

After compressing the time axis of the video sequence to the channel axis, the temporal information on the feature channel will be mixed together, and the object relationship between frames or between different frames will be more difficult to measure. Therefore, it is necessary to perform time-related recovery and enhancement on the compressed features.

In one possible implementation, processing the first feature map through the feature extraction network to obtain one or more second feature maps includes: determining the weight corresponding to the channel dimension of the input feature map based on the input feature map, and the input feature map is the intermediate output obtained by the feature extraction network processing the one or more first feature maps; performing a convolution operation on the input feature map to obtain convolution operation results of multiple channel dimensions, and fusing the convolution operation results of the multiple channel dimensions according to the weights to obtain a processing result.

In the above manner, the weights obtained by the first weight determination module are used to restore the time series information in the features, thereby enhancing the enhancement of the time series information and improving the processing performance of the network.

In one possible implementation, processing the first feature map through a feature extraction network to obtain one or more second feature maps includes: expanding an input feature map into multiple feature maps distributed along a time dimension, where the input feature map is an intermediate output obtained by the feature extraction network processing the one or more first feature maps; and performing feature interaction between the multiple feature maps obtained along the time dimension. Interaction between features, referred to as feature interaction, refers to obtaining an interaction result by calculating the relationship between different features through a certain mapping (such as convolution or attention mechanism).

In the above way, by converting the channels into frame numbers and interacting in the time dimension, the timing information is restored and the relationship between objects in the same frame or different frames is captured.

In a possible implementation, the interaction includes: interaction based on an attention mechanism, or interaction achieved through large kernel convolution.

In addition, timing information can be added through time coding. In one possible implementation,

The processing of the first feature map through a feature extraction network to obtain one or more second feature maps includes: determining a time code that is consistent with a size of the multiple feature maps; fusing the time code with the multiple feature maps to obtain multiple feature maps in a fused time dimension; and performing feature interaction between the multiple feature maps in the fused time dimension.

In a possible implementation, the interaction module is further configured to fuse the interaction result obtained through the interaction with the input feature.

In a second aspect, the present application provides a data processing device, comprising:

a compression module, configured to obtain features of a video, the features comprising a plurality of feature maps distributed in a time dimension, the dimensions of the feature maps comprising a channel dimension and a spatial dimension, and convert the plurality of feature maps from being distributed in the time dimension to being distributed in the channel dimension or the spatial dimension, to obtain the one or more first feature maps; the first feature maps do not include the time dimension;

A processing module is used to process the first feature map through a feature extraction network to obtain one or more second feature maps, and obtain a task processing result of the video through a task network based on the one or more second feature maps.

In a possible implementation, the task processing result is a processing result of a video understanding task, a video-based generation task, or a video enhancement task.

In a possible implementation, the first feature map and the feature have the same size in spatial dimension.

In one possible implementation, the size of the feature is (x, t, h, w), and the size of the first feature map is (x*t, h, w), where x is the number of channels, t is the number of times, h and w are the height and width in the spatial dimension, respectively, and * is the product.

In a possible implementation, the compression module is specifically configured to:

The feature maps at different time dimensions in the multiple feature maps are stacked in the channel dimension to obtain the one or more first feature maps.

In one possible implementation, the feature extraction network includes a first weight determination module and a convolution module;

The first weight determination module is used to determine the weight corresponding to the channel dimension of the input feature map based on the input feature map;

The convolution module is used to perform a convolution operation on the input feature map to obtain convolution operation results of multiple channel dimensions, and fuse the convolution operation results of the multiple channel dimensions according to the weights to obtain a processing result.

In one possible implementation, the feature extraction network includes a transformation module and an interaction module;

The transformation module is used to expand the input feature map into multiple feature maps distributed in the time dimension;

The interaction module is used to perform feature interaction between the multiple feature maps distributed in the time dimension.

In a possible implementation, the interaction includes:

Interaction based on attention mechanism or interaction achieved through large kernel convolution.

In a possible implementation, the transformation module is further configured to:

Determine a temporal encoding having a size consistent with the plurality of feature maps obtained by the transform module;

Fusing the multiple feature maps obtained by the time encoding and the transformation module to obtain multiple fused feature maps in the time dimension;

The interaction module is specifically used to perform feature interaction between the multiple feature maps in the fused time dimension.

In a possible implementation, the interaction module is further configured to fuse the interaction result obtained through the interaction with the input feature.

In a third aspect, an embodiment of the present application provides a data processing device, which may include a memory, a processor, and a bus system, wherein the memory is used to store programs, and the processor is used to execute the programs in the memory to perform the first aspect and any optional method thereof.

In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which a computer program is stored. When the computer-readable storage medium is run on a computer, the computer executes the above-mentioned first aspect and any optional method thereof.

In a fifth aspect, an embodiment of the present application provides a computer program, which, when executed on a computer, enables the computer to execute the above-mentioned first aspect and any optional method thereof.

In a sixth aspect, the present application provides a chip system comprising a processor configured to support the execution of a data processing device to implement the functions described in the aforementioned aspects, such as transmitting or processing the data or information described in the aforementioned methods. In one possible design, the chip system further comprises a memory configured to store program instructions and data necessary for executing the device or training the device. The chip system may consist of a single chip or may include a chip and other discrete components.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1A is a schematic diagram of a structure of an artificial intelligence main framework;

1B and 1C are schematic diagrams of the application system framework of the present invention;

FIG1D is a schematic diagram of an optional hardware structure of a terminal;

FIG2 is a schematic diagram of the structure of a server;

FIG3 is a schematic diagram of a system architecture of the present application;

Figure 4 shows a process of cloud services;

FIG5 is a flowchart of a data processing method provided in an embodiment of the present application;

FIG6 is a schematic diagram of a data processing method provided in an embodiment of the present application;

FIG7 is a schematic diagram of an effect provided by an embodiment of the present application;

FIG8 is a schematic structural diagram of a data processing device provided in an embodiment of the present application;

FIG9 is a schematic diagram of a structure of an execution device provided in an embodiment of the present application;

FIG10 is a schematic diagram of a structure of a training device provided in an embodiment of the present application;

FIG11 is a schematic diagram of the structure of a chip provided in an embodiment of the present application.

DETAILED DESCRIPTION

The following describes the embodiments of the present invention in conjunction with the accompanying drawings. The terms used in the embodiments of the present invention are only used to explain the specific embodiments of the present invention, and are not intended to limit the present invention.

The embodiments of the present application are described below in conjunction with the accompanying drawings. Those skilled in the art will appreciate that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, and this is merely a way of distinguishing the objects of the same attributes when describing them in the embodiments of the present application. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, so that the process, method, system, product or equipment comprising a series of units need not be limited to those units, but may include other units that are not clearly listed or inherent to these processes, methods, products or equipment.

As used herein, the terms "substantially," "about," and similar terms are used as terms of approximation, not as terms of degree, and are intended to take into account the inherent variations in measurements or calculations that one of ordinary skill in the art would recognize. Furthermore, the use of "may" when describing embodiments of the present invention refers to "one or more possible embodiments." As used herein, the terms "use," "using," and "used" may be considered synonymous with the terms "utilize," "utilizing," and "utilized," respectively. Additionally, the term "exemplary" is intended to refer to an example or illustration.

First, let's describe the overall workflow of an AI system. See Figure 1A, which shows a schematic diagram of the main AI framework. This AI framework will be explained from two perspectives: the "intelligent information chain" (horizontal axis) and the "IT value chain" (vertical axis). The "intelligent information chain" reflects the entire process from data acquisition to processing. For example, it could be the general process of intelligent information perception, intelligent information representation and formation, intelligent reasoning, intelligent decision-making, and intelligent execution and output. Throughout this process, data undergoes a condensed journey from "data-information-knowledge-wisdom." The "IT value chain," spanning the underlying infrastructure of human intelligence, information (provided and processed by technology), and the system's industrial ecosystem, reflects the value that AI brings to the information technology industry.

(1) Infrastructure

Infrastructure provides computing power for AI systems, enabling communication with the outside world and supporting this through a foundational platform. External communication occurs through sensors; computing power is provided by intelligent chips (CPUs, NPUs, GPUs, ASICs, FPGAs, and other hardware accelerators). The foundational platform includes a distributed computing framework and network-related platform guarantees and support, including cloud storage and computing, and interconnected networks. For example, sensors communicate with the outside world to acquire data, which is then fed into the intelligent chips within the distributed computing system provided by the foundational platform for computation.

(2) Data

Data above the infrastructure layer represents data sources for AI. This data includes graphics, images, voice, and text, as well as IoT data from traditional devices. This includes business data from existing systems and sensor data such as force, displacement, liquid level, temperature, and humidity.

(3) Data processing

Data processing generally includes data training, machine learning, deep learning, search, reasoning, decision-making, etc.

Among them, machine learning and deep learning can symbolize and formalize data for intelligent information modeling, extraction, preprocessing, and training.

Reasoning refers to the process of simulating human intelligent reasoning in computers or intelligent systems, using formalized information to perform machine thinking and solve problems based on reasoning control strategies. Typical functions are search and matching.

Decision-making refers to the process of making decisions after intelligent information is reasoned, and usually provides functions such as classification, sorting, and prediction.

(4) General ability

After the data has undergone the data processing mentioned above, some general capabilities can be further formed based on the results of the data processing, such as algorithms or a general system, for example, translation, text analysis, computer vision processing, speech recognition, image recognition, etc.

(5) Smart products and industry applications

Smart products and industry applications refer to the products and applications of artificial intelligence systems in various fields. They are the encapsulation of the overall artificial intelligence solution, which productizes intelligent information decision-making and realizes practical application. Its application areas mainly include: smart terminals, smart transportation, smart medical care, autonomous driving, smart cities, etc.

First, we will introduce the application scenarios of this application. This application can be applied to, but is not limited to, applications with video processing functions (hereinafter referred to as video processing applications) or cloud services provided by cloud-side servers. The following are introduced respectively:

Video processing can be, but is not limited to, video understanding, video enhancement, and video-based generation tasks.

Among them, video understanding can include but is not limited to: action recognition, temporal action localization, video summarization, video detection, video segmentation, multimodal video understanding, pedestrian re-identification and other tasks.

1. Video processing applications

The product form of the embodiment of the present application can be a video processing application. The video processing application can be run on a terminal device or a cloud-side server.

Among them, the video processing task in the embodiment of the present application can be: obtaining a task processing result of the video based on the video input by the user.

In a possible implementation, a video processing application may implement a video processing task based on a video input by a user and obtain a task processing result of the video.

In one possible implementation, the user can open a video processing application installed on the terminal device and input a video. The video processing application can process the video input by the user through a model trained by the method provided in the embodiment of the present application, or through the method provided in the embodiment of the present application, and present the task processing results of the video to the user (the presentation method can be but is not limited to display, playback, saving, uploading to the cloud side, etc.).

In one possible implementation, a user can open a video processing application installed on a terminal device and input a video. The video processing application can send the video to a cloud-side server. The cloud-side server processes the video using a model trained using the method provided in an embodiment of the present application, and transmits the task processing results of the video back to the terminal device. The terminal device can present the task processing results of the video to the user (the presentation method can be, but is not limited to, display, playback, saving, uploading to the cloud side, etc.).

Next, the video processing application in the embodiment of this application is introduced from the perspective of functional architecture and product architecture that implements the functions.

Referring to FIG. 1B , FIG. 1B is a schematic diagram of the functional architecture of a video processing application in an embodiment of the present application:

In one possible implementation, as shown in FIG1B , a video processing application 102 may receive input parameters 101 (e.g., including a video) and generate a video task processing result 103. The video processing application 102 may be executed on, for example, at least one computer system and include computer code that, when executed by one or more computers, causes the computers to execute a model trained using the method provided in the embodiments of the present application.

Referring to FIG. 1C , FIG. 1C is a schematic diagram of the physical architecture for running a video processing application in an embodiment of the present application:

Referring to FIG1C , FIG1C shows a schematic diagram of a system architecture. The system may include a terminal 100 and a server 200. The server 200 may include one or more servers (FIG1C illustrates one server as an example), and the server 200 may provide video processing or natural language generation functions for one or more terminals.

Among them, the terminal 100 can be installed with a video processing application, or a web page related to the video processing or natural language generation function can be opened. The above application and web page can provide an interface. The terminal 100 can receive the relevant parameters entered by the user on the video processing or natural language generation function interface, and send the above parameters to the server 200. The server 200 can obtain the task processing results of the video based on the received parameters, and return the task processing results of the video to the terminal 100.

It should be understood that in some optional implementations, the terminal 100 can also complete the action of obtaining the task processing result of the video based on the received parameters by itself without the need for cooperation from the server, and the embodiments of the present application are not limited thereto.

Next, the product form of the terminal 100 in FIG1C is described;

The terminal 100 in the embodiment of the present application can be a mobile phone, a tablet computer, a wearable device, a vehicle-mounted device, an augmented reality (AR)/virtual reality (VR) device, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), etc., and the embodiment of the present application does not impose any restrictions on this.

FIG1D shows a schematic diagram of an optional hardware structure of the terminal 100 .

1D , the terminal 100 may include components such as a radio frequency unit 110, a memory 120, an input unit 130, a display unit 140, a camera 150 (optional), an audio circuit 160 (optional), a speaker 161 (optional), a microphone 162 (optional), a processor 170, an external interface 180, and a power supply 190. Those skilled in the art will appreciate that FIG1D is merely an example of a terminal or multi-function device and does not limit the terminal or multi-function device. The terminal or multi-function device may include more or fewer components than shown, or may combine certain components or have different components.

The input unit 130 can be used to receive input digital or character information and generate key signal input related to user settings and function control of the portable multifunction device. Specifically, the input unit 130 may include a touch screen 131 (optional) and/or other input devices 132. The touch screen 131 can detect user touch operations on or near it (for example, operations performed on or near the touch screen using a finger, joint, stylus, or any other suitable object) and drive corresponding connected devices according to pre-set programs. The touch screen can detect user touch actions on the touch screen, convert the touch actions into touch signals and transmit them to the processor 170. It can also receive and execute commands sent by the processor 170; the touch signals include at least touch point coordinate information. The touch screen 131 provides an input interface and an output interface between the terminal 100 and the user. Touch screens can be implemented using various types, including resistive, capacitive, infrared, and surface acoustic wave. In addition to the touch screen 131, the input unit 130 may also include other input devices. Specifically, the other input devices 132 may include, but are not limited to, one or more of a physical keyboard, function keys (such as a volume control key, a switch key, etc.), a trackball, a mouse, a joystick, and the like.

Among them, other input devices 132 can receive input video.

The display unit 140 may be used to display information input by the user or provided to the user, various menus of the terminal 100, interactive interfaces, file display, and/or playback of any multimedia file. In an embodiment of the present application, the display unit 140 may be used to display the interface of a generated application, processing results, etc.

Memory 120 can be used to store instructions and data. It primarily includes an instruction storage area and a data storage area. The data storage area can store various data, such as multimedia files and text. The instruction storage area can store software units such as the operating system, applications, and instructions required for at least one function, or subsets or extensions thereof. It may also include non-volatile random access memory (RAM). It provides processor 170 with management functions for the hardware, software, and data resources within the computing and processing device, supporting control software and applications. It is also used to store multimedia files and running programs and applications.

The processor 170 is the control center of the terminal 100. It connects all components of the terminal 100 using various interfaces and circuits. By executing instructions stored in the memory 120 and accessing data stored therein, it executes various functions of the terminal 100 and processes data, thereby providing overall control of the terminal device. Optionally, the processor 170 may include one or more processing units. Preferably, the processor 170 may integrate an application processor and a modem processor, with the application processor primarily processing the operating system, user interface, and application programs, while the modem processor primarily handles wireless communications. It is understood that the modem processor may not be integrated into the processor 170. In some embodiments, the processor and memory may be implemented on a single chip; in other embodiments, they may be implemented on separate chips. The processor 170 may also generate corresponding operational control signals and send them to the corresponding components of the computing and processing device. It may also read and process data in the software, particularly the data and programs in the memory 120, to enable the various functional modules therein to perform their corresponding functions, thereby controlling the corresponding components to operate as instructed.

Among them, the memory 120 can be used to store software codes related to the data processing method, the processor 170 can execute the steps of the chip's data processing method, and can also schedule other units (such as the above-mentioned input unit 130 and display unit 140) to achieve corresponding functions.

The RF unit 110 (optional) can be used to send and receive information or receive and send signals during a call. For example, it receives downlink information from the base station and sends it to the processor 170 for processing; in addition, it sends uplink data to the base station. Typically, the RF circuit includes, but is not limited to, an antenna, at least one amplifier, a transceiver, a coupler, a low-noise amplifier (LNA), a duplexer, etc. In addition, the RF unit 110 can also communicate with network devices and other devices via wireless communication. This wireless communication can use any communication standard or protocol, including but not limited to Global System of Mobile Communications (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), email, Short Messaging Service (SMS), etc.

In this embodiment of the present application, the RF unit 110 can send the video to the server 200 and receive the processing result sent by the server 200.

It should be understood that the radio frequency unit 110 is optional and can be replaced by other communication interfaces, such as a network port.

The terminal 100 also includes a power supply 190 (such as a battery) for supplying power to various components. Preferably, the power supply can be logically connected to the processor 170 through a power management system, thereby managing functions such as charging, discharging, and power consumption through the power management system.

The terminal 100 also includes an external interface 180, which can be a standard Micro USB interface or a multi-pin connector. It can be used to connect the terminal 100 to communicate with other devices, and can also be used to connect a charger to charge the terminal 100.

Although not shown, terminal 100 may also include a flashlight, a wireless fidelity (WiFi) module, a Bluetooth module, sensors with different functions, etc., which are not described in detail here. Some or all of the methods described below may be applied to terminal 100 as shown in FIG. 1D .

Next, the product form of the server 200 in FIG1C is described;

FIG2 provides a schematic diagram of the structure of a server 200. As shown in FIG2, the server 200 includes a bus 201, a processor 202, a communication interface 203, and a memory 204. The processor 202, the memory 204, and the communication interface 203 communicate with each other via the bus 201.

Bus 201 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. Buses can be classified as address buses, data buses, control buses, and the like. For ease of illustration, FIG2 shows only one thick line, but this does not imply that there is only one bus or only one type of bus.

The processor 202 can be any one or more of a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP).

The memory 204 may include volatile memory, such as random access memory (RAM). The memory 204 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard drive (HDD), or solid state drive (SSD).

The memory 204 may be used to store software codes related to the data processing method, and the processor 202 may execute the steps of the data processing method of the chip, and may also schedule other units to implement corresponding functions.

It should be understood that the above-mentioned terminal 100 and server 200 can be centralized or distributed devices, and the processors in the above-mentioned terminal 100 and server 200 (such as processor 170 and processor 202) can be hardware circuits (such as application specific integrated circuit (ASIC), field-programmable gate array (FPGA), general-purpose processor, digital signal processor (DSP), microprocessor or microcontroller, etc.), or a combination of these hardware circuits. For example, the processor can be a hardware system with the function of executing instructions, such as CPU, DSP, etc., or a hardware system without the function of executing instructions, such as ASIC, FPGA, etc., or a combination of the above-mentioned hardware systems without the function of executing instructions and hardware systems with the function of executing instructions.

It should be understood that the steps related to the model reasoning process in the embodiments of this application involve AI-related operations. When performing AI operations, the instruction execution architecture of the terminal device and server is not limited to the processor-memory architecture described above. The system architecture provided in the embodiments of this application is described in detail below with reference to Figure 3.

FIG3 is a schematic diagram of the system architecture provided by an embodiment of the present application. As shown in FIG3 , the system architecture 500 includes an execution device 510 , a training device 520 , a database 530 , a client device 540 , a data storage system 550 , and a data acquisition system 560 .

The execution device 510 includes a calculation module 511, an I/O interface 512, a pre-processing module 513, and a post-processing module 514. The calculation module 511 may include the target model/rule 501, and the pre-processing module 513 and the post-processing module 514 are optional.

The execution device 510 may be a terminal device or a server that runs the aforementioned generated application program.

The data acquisition device 560 is used to collect training samples. The training samples can be image data, etc. After collecting the training samples, the data acquisition device 560 stores these training samples in the database 530.

The training device 520 can train the neural network to be trained (such as the feature extraction network, task network and other neural networks in the embodiments of the present application) based on the training samples maintained in the database 530 to obtain the target model/rule 501.

It should be understood that the training device 520 can perform a pre-training process on the neural network to be trained based on the training samples maintained in the database 530, or fine-tune the model based on the pre-training.

It should be noted that, in actual applications, the training samples maintained in the database 530 may not all be collected by the data acquisition device 560, but may also be received from other devices. It should also be noted that the training device 520 may not train the target model/rule 501 entirely based on the training samples maintained in the database 530, but may also obtain training samples from the cloud or other places for model training. The above description should not be used as a limitation on the embodiments of the present application.

The target model/rule 501 obtained through training with the training device 520 can be applied to different systems or devices, such as the execution device 510 shown in FIG3 . The execution device 510 can be a terminal, such as a mobile phone terminal, a tablet computer, a laptop computer, an augmented reality (AR)/virtual reality (VR) device, a vehicle-mounted terminal, etc., or a server, etc.

Specifically, the training device 520 may transfer the trained model to the execution device 510 .

In Figure 3, the execution device 510 is configured with an input/output (I/O) interface 512 for data interaction with external devices. The user can input data (such as video in the embodiment of the present application) to the I/O interface 512 through the client device 540.

Preprocessing module 513 and preprocessing module 514 are used to preprocess the input data received by I/O interface 512. It should be understood that preprocessing module 513 and preprocessing module 514 may be absent or only one preprocessing module may be present. If preprocessing module 513 and preprocessing module 514 are absent, computing module 511 may be used directly to process the input data.

When the execution device 510 preprocesses the input data, or when the computing module 511 of the execution device 510 performs calculations and other related processing, the execution device 510 can call the data, code, etc. in the data storage system 550 for corresponding processing, and can also store the data, instructions, etc. obtained from the corresponding processing in the data storage system 550.

Finally, the I/O interface 512 provides the processed results to the client device 540 and thus to the user.

In the scenario shown in FIG3 , the user can manually input data, and this "manual input data" can be operated through the interface provided by I/O interface 512. In another scenario, client device 540 can automatically send input data to I/O interface 512. If user authorization is required for client device 540 to automatically send input data, the user can set the corresponding permissions in client device 540. The user can view the results output by execution device 510 on client device 540, and the specific presentation form can be a display, sound, action, or other specific method. Client device 540 can also serve as a data acquisition terminal, collecting input data input into I/O interface 512 and output results output from I/O interface 512 as new sample data, and storing them in database 530. Of course, collection can also be performed without client device 540, and instead the I/O interface 512 directly stores the input data input into I/O interface 512 and output results output from I/O interface 512 as new sample data in database 530.

It is worth noting that FIG3 is merely a schematic diagram of a system architecture provided by an embodiment of the present application, and the positional relationships between the devices, components, modules, etc. shown in the figure do not constitute any limitation. For example, in FIG3 , the data storage system 550 is an external memory relative to the execution device 510. In other cases, the data storage system 550 can also be placed in the execution device 510. It should be understood that the execution device 510 can be deployed in the client device 540.

From the inference side of the model:

In the embodiment of the present application, the computing module 511 of the above-mentioned execution device 510 can obtain the code stored in the data storage system 550 to implement the steps related to the model reasoning process in the embodiment of the present application.

In an embodiment of the present application, the computing module 511 of the execution device 510 may include a hardware circuit (such as an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a general-purpose processor, a digital signal processor (DSP), a microprocessor or a microcontroller, etc.), or a combination of these hardware circuits. For example, the training device 520 may be a hardware system with an instruction execution function, such as a CPU, DSP, etc., or a hardware system without an instruction execution function, such as an ASIC, FPGA, etc., or a combination of the above-mentioned hardware systems without an instruction execution function and hardware systems with an instruction execution function.

Specifically, the computing module 511 of the execution device 510 can be a hardware system with an execution instruction function, and the steps related to the model reasoning process provided in the embodiment of the present application can be software codes stored in the memory. The computing module 511 of the execution device 510 can obtain the software code from the memory and execute the obtained software code to implement the steps related to the model reasoning process provided in the embodiment of the present application.

It should be understood that the computing module 511 of the execution device 510 can be a combination of a hardware system that does not have the function of executing instructions and a hardware system that has the function of executing instructions. Some of the steps related to the model reasoning process provided in the embodiment of the present application can also be implemented by the hardware system that does not have the function of executing instructions in the computing module 511 of the execution device 510, which is not limited here.

From the training side of the model:

In an embodiment of the present application, the above-mentioned training device 520 can obtain the code stored in the memory (not shown in Figure 3, which can be integrated into the training device 520 or deployed separately from the training device 520) to implement the steps related to model training in the embodiment of the present application.

In an embodiment of the present application, the training device 520 may include a hardware circuit (such as an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a general-purpose processor, a digital signal processor (DSP), a microprocessor or a microcontroller, etc.), or a combination of these hardware circuits. For example, the training device 520 may be a hardware system with an instruction execution function, such as a CPU, DSP, etc., or a hardware system without an instruction execution function, such as an ASIC, FPGA, etc., or a combination of the above-mentioned hardware systems without an instruction execution function and hardware systems with an instruction execution function.

It should be understood that the training device 520 can be a combination of a hardware system that does not have the function of executing instructions and a hardware system that has the function of executing instructions. Some of the steps related to model training provided in the embodiments of the present application can also be implemented by the hardware system in the training device 520 that does not have the function of executing instructions, which is not limited here.

3. Video processing cloud services provided by the server:

In one possible implementation, the server can provide video processing service to the end side through an application programming interface (API).

Among them, the terminal device can send relevant parameters (such as video) to the server through the API provided by the cloud. The server can obtain processing results based on the received parameters, etc., and return the processing results to the terminal.

The description of the terminal and the server can be the same as that of the above embodiments, and will not be repeated here.

FIG4 shows a process of using a video processing function cloud service provided by a cloud platform.

1. Activate and purchase video processing services.

2. Users can download the software development kit (SDK) corresponding to the video processing service. Usually, the cloud platform provides multiple development versions of the SDK for users to choose according to the requirements of the development environment, such as JAVA version SDK, Python version SDK, PHP version SDK, Android version SDK, etc.

3. After the user downloads the corresponding version of the SDK to the local computer as needed, import the SDK project into the local development environment, configure and debug it in the local development environment. The local development environment can also be used to develop other functions, forming an application that integrates video processing functional capabilities.

4. When a video processing application is used and needs to perform video processing, it can trigger an API call for that function. When the application triggers the video processing function, it initiates an API request to the running instance of the video processing service in the cloud environment. The API request includes image data, and the running instance in the cloud environment processes the image to obtain the processing result.

5. The cloud environment returns the processing results to the application, thus completing a video processing function call.

Since the embodiments of the present application involve the application of a large number of neural networks, in order to facilitate understanding, the relevant terms and related concepts such as neural networks involved in the embodiments of the present application are first introduced below.

(1) Neural Network

A neural network can be composed of neural units. A neural unit can refer to an operation unit that takes xs (i.e., input data) and intercept 1 as input. The output of the operation unit can be:

Where s = 1, 2, ... n, where n is a natural number greater than 1, Ws is the weight of xs, and b is the bias of the neural unit. f is the activation function of the neural unit, which is used to introduce nonlinear characteristics into the neural network to convert the input signal of the neural unit into the output signal. The output signal of this activation function can be used as the input of the next convolutional layer, and the activation function can be a sigmoid function. A neural network is a network formed by connecting multiple single neural units mentioned above, that is, the output of one neural unit can be the input of another neural unit. The input of each neural unit can be connected to the local receptive field of the previous layer to extract the features of the local receptive field. The local receptive field can be an area composed of several neural units.

(2) Convolutional neural network (CNN) is a deep neural network with a convolutional structure. A convolutional neural network contains a feature extractor consisting of a convolution layer and a subsampling layer, which can be regarded as a filter. A convolution layer refers to a neuron layer in a convolutional neural network that performs convolution processing on the input signal. In the convolution layer of a convolutional neural network, a neuron can only be connected to some neurons in the adjacent layers. A convolution layer usually contains several feature planes, and each feature plane can be composed of a number of rectangularly arranged neural units. The neural units in the same feature plane share weights, and the shared weights here are convolution kernels. Shared weights can be understood as the way of extracting features is independent of position. The convolution kernel can be formalized as a matrix of random size, and the convolution kernel can obtain reasonable weights through learning during the training process of the convolutional neural network. In addition, the direct benefit of shared weights is to reduce the connections between the layers of the convolutional neural network, while reducing the risk of overfitting.

CNN is a very common neural network. The following will focus on a detailed introduction to its structure. As mentioned in the previous basic concepts, a convolutional neural network is a deep neural network with a convolutional structure. It is a deep learning architecture, which uses machine learning algorithms to perform multiple levels of learning at different levels of abstraction. As a deep learning architecture, CNN is a feed-forward artificial neural network, in which each neuron responds to an image input.

(3) Deep Neural Networks

Deep Neural Network (DNN), also known as multi-layer neural network, can be understood as a neural network with many hidden layers. There is no special metric for "many" here. Based on the position of different layers in DNN, the neural network inside DNN can be divided into three categories: input layer, hidden layer, and output layer. Generally speaking, the first layer is the input layer, the last layer is the output layer, and the layers in between are all hidden layers. The layers are fully connected, that is, any neuron in the i-th layer must be connected to any neuron in the i+1-th layer. Although DNN looks complicated, the work of each layer is actually not complicated. Simply put, it is the following linear relationship expression: in, is the input vector, is the output vector, is the offset vector, W is the weight matrix (also called coefficient), and α() is the activation function. Each layer is just an input vector After such a simple operation, the output vector Since there are many DNN layers, the coefficient W and the offset vector The definition of these parameters in DNN is as follows: Take the coefficient W as an example: Assume that in a three-layer DNN, the linear coefficient from the 4th neuron in the second layer to the 2nd neuron in the third layer is defined as The superscript 3 represents the layer number of the coefficient W, while the subscript corresponds to the output of the third layer index 2 and the input of the second layer index 4. In summary, the coefficient from the kth neuron in the L-1th layer to the jth neuron in the Lth layer is defined as It's important to note that the input layer has no W parameter. In deep neural networks, more hidden layers allow the network to better capture complex real-world situations. Theoretically, a model with more parameters has higher complexity and greater "capacity," meaning it can handle more complex learning tasks. Training a deep neural network is essentially the process of learning the weight matrix, with the ultimate goal of obtaining the weight matrices for all layers of a trained deep neural network (a weight matrix formed by the vectors W across many layers).

(4) Loss function

When training a deep neural network, we want the network's output to be as close as possible to the desired predicted value. This is done by comparing the network's predictions with the desired target values and then updating the weight vectors of each layer based on the difference between the two. (Of course, before the first update, there's usually an initialization process, which pre-configures the parameters for each layer in the deep neural network.) For example, if the network's prediction is too high, the weight vectors are adjusted to predict a lower value. This adjustment is repeated until the deep neural network can predict the desired target value or a value very close to it. Therefore, it's necessary to predefine how to compare the difference between the predicted and target values. This is known as the loss function, or objective function, a crucial equation used to measure the difference between the predicted and target values. For example, a higher loss function indicates a greater difference, so training a deep neural network becomes a process of minimizing this loss.

(5) Backpropagation algorithm

The back propagation (BP) algorithm can be used to correct the size of the initial model parameters during training, reducing the model's error loss. Specifically, forward propagation of the input signal to the output generates error loss. This error loss information is then backpropagated to update the parameters in the initial model, thereby converging the error loss. The BP algorithm is a backpropagation algorithm driven by error loss, aiming to obtain optimal model parameters, such as the weight matrix.

Unlike image recognition, video processing tasks (such as video understanding) require more computing resources. Therefore, it is crucial to research lightweight video understanding models to reduce resource consumption. Lightweight video understanding models have many practical applications, such as autonomous driving, robotics, and industrial control. However, existing video understanding models focus more on achieving higher accuracy, resulting in relatively large model designs, while paying little attention to improving the performance of small models in end-to-end applications.

Therefore, there is an urgent need for a lightweight model design for video processing.

In existing technologies, whether convolution-based or transformer-based methods, they all treat the timeline of the video sequence as a separate dimension for video processing. This requires a lot of additional computing and memory usage to process information in the time dimension, resulting in high resource consumption and a lack of significant speed advantage, making it difficult to apply to mobile devices. This embodiment of the present application proposes a lightweight video understanding model based on timeline compression and temporal feature recovery, which can be applied to video understanding tasks in mobile and end-side applications with superior performance.

The present invention provides a data processing method. The data processing method of the present invention is described in detail below with reference to the accompanying drawings.

Refer to Figure 5, which is a flow chart of a data processing method provided in an embodiment of the present application. As shown in Figure 5, a data processing method provided in an embodiment of the present application may include steps 501 to 503, and these steps are described in detail below.

501. Obtain features of a video, where the features include multiple feature maps distributed in a time dimension, and the dimensions of the feature maps include a channel dimension and a spatial dimension.

The features of a video can be obtained by extracting features from the video. Since the video itself includes data distributed along the time dimension, that is, multiple time frames, each frame corresponding to a frame of image data, the features extracted from the video can include multiple feature maps distributed along the time channel, each feature map including channel dimensions and spatial dimensions. In other words, the features of a video are at least four-dimensional features. Too high a dimensionality will result in excessive computational overhead for subsequent operations.

502. Change the distribution of the multiple feature maps in the time dimension to a distribution in the channel dimension or the spatial dimension to obtain the one or more first feature maps.

Existing video understanding models have poor performance on mobile terminals, while models with better performance consume too much resources, which is not conducive to application to end-side devices. It is difficult to balance the resource consumption and performance of mobile terminal models. The idea of this application is to reduce the dimension of video features. Specifically, the features in the time dimension are compressed to the channel dimension or the spatial dimension, so that the compressed features (that is, the first feature map in the embodiment of this application) do not include the time dimension, and the amount of data in the channel dimension and the spatial dimension increases. When the feature dimension is reduced, the amount of subsequent calculations required can be greatly reduced, for example, it can be changed from 3D convolution to 2D convolution.

Taking compression to the channel dimension as an example, in one possible implementation, the size of the feature is (x, t, h, w), and the size of the first feature map is (x*t, h, w), where x is the number of channels, t is the number of times, h and w are the height and width in the spatial dimension, respectively, and * is the product.

In a possible implementation, the first feature map and the feature have the same size in the spatial dimension, while the size of the first feature map in the channel dimension is larger than the size of the video feature in the channel dimension.

In a possible implementation, the feature maps in different time dimensions among the multiple feature maps may be stacked in the channel dimension to obtain the one or more first feature maps, and the stacking order may be based on the order in the time dimension.

Taking compression to the channel dimension as an example, in a possible implementation, the feature maps in different time dimensions among the multiple feature maps can be stacked in the spatial dimension to obtain the one or more first feature maps, and the first feature map and the feature have the same size in the channel dimension, while the size of the first feature map in the spatial dimension is larger than the size of the video feature in the channel dimension.

Alternatively, the time dimension in the video features can be compressed to the channel dimension and the spatial dimension at the same time. In this case, the size of the first feature map in the spatial dimension is larger than the size of the video features in the channel dimension, and the size of the first feature map in the spatial dimension is larger than the size of the video features in the channel dimension.

As shown in FIG6 , FIG6 is an overall block diagram of a lightweight video understanding model based on time axis compression and time feature enhancement proposed in an embodiment of the present application. The overall network process is as follows: (1) For an input video sequence X, whose vector shape is (3, t, h, w), its time axis is compressed to the channel axis through a vector deformation operation (which can reduce the large amount of computing resources consumed in subsequent feature processing), that is, an input with a shape of (3t, h, w) is obtained.

503. Obtain a task processing result of the video according to the one or more first feature maps.

In a possible implementation, the first feature map can be processed by a feature extraction network to obtain one or more second feature maps, and the task processing result of the video can be obtained through a task network based on the one or more second feature maps.

After compressing the time axis of the video sequence to the channel axis, the temporal information on the feature channel will be mixed together, and the object relationship between frames or between different frames will be more difficult to measure. Therefore, it is necessary to perform time-related recovery and enhancement on the compressed features.

In one possible implementation, the feature extraction network may include, but is not limited to, a convolutional or transformer network. For example, the feature extraction network may include multiple feature extraction units, each of which is connected via a downsampling operation, and each of which is used to process features at different scales.

In the embodiment of the present application, in order to perform time-dependent recovery and enhancement on the compressed features (first feature map), relevant operations are added to the feature extraction network, which are described below:

In a possible implementation, the feature extraction network may include a first weight determination module and a convolution module. The first weight determination module and the convolution module may belong to the at least one feature extraction unit.

The first weight determination module is used to determine the weight corresponding to the channel dimension of the input feature map based on the input feature map. The weight can represent the temporal importance of the channel, and the weights corresponding to different channels can be different. The convolution module is used to perform a convolution operation on the input feature map to obtain convolution operation results of multiple channel dimensions, and fuse the convolution operation results of the multiple channel dimensions according to the weight to obtain a processing result. In this way, the weights obtained by the first weight determination module are used to restore the timing information in the feature, thereby enhancing the enhancement of the timing information and improving the processing performance of the network.

Exemplarily, as shown in Figure 6, the above-mentioned first weight determination module and convolution module can be the time importance branch in the time channel learning unit. The time channel learning unit consists of two branches, one branch is the time importance branch, and the other branch is the cross-time object interaction module. The outputs of the two branches are fused together through a summation operation to obtain the final output.

The temporal importance learning branch is used to capture the temporal importance of the channel. It consists of a 1×1 temporal attention convolution. The formula of the temporal attention convolution is as follows:

Among them, f(x, y) represents the output value of 2D convolution at the point (x, y), k is the size of the convolution kernel, c is the number of channels, g represents the convolution kernel, h represents the feature map, the convolution operation can be implemented based on the convolution module introduced in the above embodiment, _wm is the input adaptive weight calculated according to the input feature (that is, the weight corresponding to the channel dimension of the input feature map introduced above), _wm can be obtained in various ways, and can be obtained through a multi-layer perceptron, a global attention module, etc. (which can be implemented based on the first weight determination module introduced in the above embodiment).

In one possible implementation, the feature extraction network may include a transformation module and an interaction module. The transformation module and the interaction module may belong to at least one of the aforementioned feature extraction units. In particular, the transformation module and the interaction module may belong to the same feature extraction unit as the first weight determination module and the convolution module described above.

For example, the result obtained by the interaction module (or the result obtained by performing other processing based on the result obtained by the interaction module) can be further integrated with the processing result obtained by the first weight determination module.

In one possible implementation, the transformation module is used to expand the input feature map into multiple feature maps distributed in the time dimension. For example, the input feature map can be expanded into multiple feature maps distributed in the time dimension through a channel conversion function. For example, the number of expanded features in the time dimension can be consistent with the number of video features in the time dimension.

The interaction module is used to perform feature interaction between the multiple feature maps distributed in the time dimension.

For example, in one possible implementation, the interaction includes: interaction based on an attention mechanism, or interaction achieved through large kernel convolution.

In one possible implementation, the transformation module is further configured to determine a time code that is consistent in size with the multiple feature maps obtained by the transformation module. The time code may be a four-dimensional tensor. Each eigenvalue of the multiple feature maps obtained by the transformation module may correspond to a code value in the time code. The time code and the multiple feature maps obtained by the transformation module may be fused to obtain multiple feature maps in a fused time dimension. The fusion may be an addition operation at corresponding positions. Furthermore, the interaction module is specifically configured to perform feature interaction between the multiple feature maps in the fused time dimension.

In addition, residual connections can be added. In one possible implementation, the interaction module is also used to fuse the interaction results obtained through the interaction and the input features. For example, the interaction results obtained through the interaction and the input features can be first mapped to information of the same size, and then the corresponding positions are added to perform fusion.

The transformation module and interaction module described above can be referred to as the cross-temporal object interaction module. The cross-temporal object interaction module can be used to restore temporal information and capture object relationships between different frames. Figure 6 (c) shows a schematic diagram of the cross-temporal object interaction module, and its functionality is described by the following formula:

Among them, _Fb represents the input feature, _FT is the time code, Is the channel conversion function, which means converting the number of channels from _Ci to the number T. It is a cross-time function used to learn the relationship between objects in different frames. φ represents the Sigmoid activation function. Represents element-wise multiplication. Among them, the channel conversion function can be implemented by 2D convolution; the time code _FT is a learnable/unlearnable position code; the cross-time function It can be implemented by various modules that capture global positional relationships, for example, it can be implemented by large-kernel 2D convolution or by a windowed global attention mechanism.

The cross-temporal object interaction module is used to restore temporal information and capture object relationships within or across frames. This module converts channels into frame numbers, adds temporal information through temporal position encoding, captures object relationships across frames using a cross-temporal function, and finally converts the channel number into the output channel number.

Referring to Figure 6, the input can be processed by a 5×5 convolution with a stride of 2, and then the input features are obtained through a maximum pooling module; the input features are input to the subsequent network of four stages, each stage contains a different number of temporal channel learning unit (CTL) blocks, each stage processes features of different scales, and the downsampling operation is used to obtain the downsampled features of each next stage; for the output features of the last stage, an average pooling and fully connected layer are input to obtain the final output of the network.

This framework only requires 2D convolutions to process videos. The core module is the CTL block, which performs the main function of temporal feature enhancement. The temporal learning unit block can be a residual block, which includes a 1×1 convolution to reduce the number of input channels to 1/4 of the original number, a core temporal channel learning unit (CTL), and another 1×1 convolution to expand the number of input channels by 4 times, returning to the original number of input channels.

Through the above methods, this application simultaneously proposes a temporal feature enhancement strategy of temporal importance learning and cross-temporal object interaction to enhance compressed feature learning and improve the performance of the overall network.

Next, the beneficial effects of the embodiments of the present application are introduced in combination with experiments. Sufficient experiments are carried out on the Kinetcs400 dataset to verify the effectiveness of the present invention. As shown in Figure 7, the two-branch structure of the time channel learning unit in the embodiment of the present application has the best performance.

In addition, the temporal channel learning unit and cross-temporal object interaction module proposed in the present invention can also be embedded in ordinary 2D convolutional networks. They are not limited to video understanding tasks. For example, they can be applied to ordinary 2D backbone networks to improve network performance.

8 , which is a schematic diagram of the structure of a data processing device provided in an embodiment of the present application. As shown in FIG8 , a data processing device provided in an embodiment of the present application, the device 800 includes:

An acquisition module 801 is configured to acquire features of a video, wherein the features include multiple feature maps distributed along a time dimension, wherein the dimensions of the feature maps include a channel dimension and a spatial dimension, and to convert the multiple feature maps from being distributed along the time dimension to being distributed along the channel dimension or the spatial dimension, thereby obtaining one or more first feature maps; wherein the first feature maps do not include the time dimension.

For a detailed description of the acquisition module 801 , reference may be made to the description of steps 501 and 502 in the above embodiment, which will not be repeated here.

The processing module 802 is used to process the first feature map through a feature extraction network to obtain one or more second feature maps, and obtain a task processing result of the video through a task network based on the one or more second feature maps.

For a detailed description of the processing module 802 , reference may be made to the description of step 503 in the above embodiment, which will not be repeated here.

In a possible implementation, the task processing result is a processing result of a video understanding task, a video-based generation task, or a video enhancement task.

In a possible implementation, the first feature map and the feature have the same size in spatial dimension.

In one possible implementation, the size of the feature is (x, t, h, w), and the size of the first feature map is (x*t, h, w), where x is the number of channels, t is the number of times, h and w are the height and width in the spatial dimension, respectively, and * is the product.

In a possible implementation, the compression module is specifically configured to:

The feature maps at different time dimensions in the multiple feature maps are stacked in the channel dimension to obtain the one or more first feature maps.

In one possible implementation, the feature extraction network includes a first weight determination module and a convolution module;

The first weight determination module is used to determine the weight corresponding to the channel dimension of the input feature map based on the input feature map;

The convolution module is used to perform a convolution operation on the input feature map to obtain convolution operation results of multiple channel dimensions, and fuse the convolution operation results of the multiple channel dimensions according to the weights to obtain a processing result.

In one possible implementation, the feature extraction network includes a transformation module and an interaction module;

The transformation module is used to expand the input feature map into multiple feature maps distributed in the time dimension;

The interaction module is used to perform feature interaction between the multiple feature maps distributed in the time dimension.

In a possible implementation, the interaction includes:

Interaction based on attention mechanism or interaction achieved through large kernel convolution.

In a possible implementation, the transformation module is further configured to:

Determine a temporal encoding having a size consistent with the plurality of feature maps obtained by the transform module;

Fusing the multiple feature maps obtained by the time encoding and the transformation module to obtain multiple fused feature maps in the time dimension;

The interaction module is specifically used to perform feature interaction between the multiple feature maps in the fused time dimension.

In a possible implementation, the interaction module is further configured to fuse the interaction result obtained through the interaction with the input feature.

Next, a terminal device provided in an embodiment of the present application is introduced. Please refer to Figure 9. Figure 9 is a structural diagram of a terminal device provided in an embodiment of the present application. The terminal device 900 can be specifically manifested as a virtual reality VR device, a mobile phone, a tablet, a laptop computer, a smart wearable device, etc., which is not limited here. Specifically, the terminal device 900 includes: a receiver 901, a transmitter 902, a processor 903 and a memory 904 (wherein the number of processors 903 in the terminal device 900 can be one or more, and Figure 9 takes one processor as an example), wherein the processor 903 may include an application processor 9031 and a communication processor 9032. In some embodiments of the present application, the receiver 901, the transmitter 902, the processor 903 and the memory 904 may be connected via a bus or other means.

The memory 904 may include read-only memory and random access memory, and provides instructions and data to the processor 903. A portion of the memory 904 may also include non-volatile random access memory (NVRAM). The memory 904 stores processor and operation instructions, executable modules, or data structures, or subsets or extensions thereof. The operation instructions may include various operation instructions for implementing various operations.

Processor 903 controls the operation of the execution device. In specific applications, the various components of the execution device are coupled together via a bus system. In addition to a data bus, the bus system may also include a power bus, a control bus, and a status signal bus. However, for clarity, all bus systems are referred to as a bus system in the figure.

The methods disclosed in the above embodiments of the present application can be applied to the processor 903 or implemented by the processor 903. The processor 903 can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by the hardware integrated logic circuit in the processor 903 or by instructions in the form of software. The above processor 903 can be a general-purpose processor, a digital signal processor (DSP), a microprocessor or a microcontroller, and can further include an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The processor 903 can implement or execute the various methods, steps and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in conjunction with the embodiments of the present application can be directly embodied as being executed by a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium well-known in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 904. Processor 903 reads information from memory 904 and, in conjunction with its hardware, completes the steps involved in the model training or model inference process in the above method.

Receiver 901 can be used to receive input digital or character information and generate signal input related to executing device-related settings and function control. Transmitter 902 can be used to output digital or character information through the first interface. Transmitter 902 can also be used to send instructions to the disk pack through the first interface to modify data in the disk pack. Transmitter 902 can also include a display device such as a display screen.

The embodiment of the present application also provides a server. Please refer to Figure 10. Figure 10 is a structural diagram of a server provided by an embodiment of the present application. The server 1000 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPU) 1010 (for example, one or more processors) and a memory 1032, and one or more storage media 1030 (for example, one or more mass storage devices) storing application programs 1042 or data 1044. Among them, the memory 1032 and the storage medium 1030 can be temporary storage or permanent storage. The program stored in the storage medium 1030 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the server. Furthermore, the central processing unit 1010 can be configured to communicate with the storage medium 1030 to execute a series of instruction operations in the storage medium 1030 on the server 1000.

The server 1000 may also include one or more power supplies 1026, one or more wired or wireless network interfaces 1050, one or more input and output interfaces 1058; or, one or more operating systems 1041, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

In an embodiment of the present application, the central processing unit 1010 is used to execute actions related to model training or model reasoning in the above embodiments.

An embodiment of the present application also provides a computer program product, which, when running on a computer, enables the computer to execute the steps executed by the aforementioned execution device, or enables the computer to execute the steps executed by the aforementioned training device.

A computer-readable storage medium is also provided in an embodiment of the present application, which stores a program for signal processing. When the computer-readable storage medium is run on a computer, it enables the computer to execute the steps executed by the aforementioned execution device, or enables the computer to execute the steps executed by the aforementioned training device.

The execution device, training device or terminal device provided in the embodiments of the present application may specifically be a chip, and the chip includes: a processing unit and a communication unit, wherein the processing unit may be, for example, a processor, and the communication unit may be, for example, an input/output interface, a pin or a circuit, etc. The processing unit may execute the computer execution instructions stored in the storage unit, so that the chip in the execution device executes the data processing method described in the above embodiment, or so that the chip in the training device executes the data processing method described in the above embodiment. Optionally, the storage unit is a storage unit in the chip, such as a register, a cache, etc. The storage unit may also be a storage unit located outside the chip in the wireless access device end, such as a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM), etc.

Specifically, see Figure 11 , which illustrates a schematic diagram of the chip structure provided in an embodiment of the present application. The chip may be a neural network processor (NPU) 1100. NPU 1100 is mounted on a host CPU (host CPU) as a coprocessor, with tasks assigned by the host CPU. The core of the NPU is arithmetic circuit 1103, which is controlled by controller 1104 to retrieve matrix data from memory and perform multiplication operations.

In some implementations, arithmetic circuit 1103 includes multiple processing elements (PEs). In some implementations, arithmetic circuit 1103 is a two-dimensional systolic array. Arithmetic circuit 1103 can also be a one-dimensional systolic array or other electronic circuitry capable of performing mathematical operations such as multiplication and addition. In some implementations, arithmetic circuit 1103 is a general-purpose matrix processor.

For example, assume there are input matrix A, weight matrix B, and output matrix C. The arithmetic circuit retrieves the corresponding data of matrix B from weight memory 1102 and caches it on each PE in the arithmetic circuit. The arithmetic circuit retrieves the data of matrix A from input memory 1101 and performs a matrix operation on matrix B. The partial or final matrix result is stored in accumulator 1108.

Unified memory 1106 is used to store input and output data. Weight data is directly transferred to weight memory 1102 via the Direct Memory Access Controller (DMAC) 1105. Input data is also transferred to unified memory 1106 via the DMAC.

BIU stands for Bus Interface Unit, that is, the bus interface unit 1110, which is used for the interaction between the AXI bus and the DMAC and instruction fetch buffer (IFB) 1109.

The bus interface unit 1110 (BIU) is used for the instruction fetch memory 1109 to obtain instructions from the external memory, and is also used for the storage unit access controller 1105 to obtain the original data of the input matrix A or the weight matrix B from the external memory.

DMAC is mainly used to move input data in the external memory DDR to the unified memory 1106 or to move weight data to the weight memory 1102 or to move input data to the input memory 1101.

The vector calculation unit 1107 includes multiple processing units. When necessary, it further processes the output of the calculation circuit 1103, such as vector multiplication, vector addition, exponential operation, logarithmic operation, size comparison, etc. It is mainly used for non-convolutional/fully connected layer network calculations in neural networks, such as batch normalization, pixel-level summation, and upsampling of feature planes.

In some implementations, the vector calculation unit 1107 can store the processed output vector in the unified memory 1106. For example, the vector calculation unit 1107 can apply a linear function or a nonlinear function to the output of the operation circuit 1103, such as linear interpolation of the feature plane extracted by the convolution layer, or accumulate a vector of values to generate an activation value. In some implementations, the vector calculation unit 1107 generates a normalized value, a pixel-level summed value, or both. In some implementations, the processed output vector can be used as an activation input to the operation circuit 1103, for example, for use in subsequent layers in a neural network.

An instruction fetch buffer 1109 connected to the controller 1104 is used to store instructions used by the controller 1104;

Unified memory 1106, input memory 1101, weight memory 1102, and instruction fetch memory 1109 are all on-chip memories. External memories are private to the NPU hardware architecture.

The processor mentioned in any of the above places can be a general-purpose central processing unit, a microprocessor, an ASIC, or one or more integrated circuits for controlling the execution of the above program.

It should also be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided in this application, the connection relationship between the modules indicates that there is a communication connection between them, which can be specifically implemented as one or more communication buses or signal lines.

Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course can also be implemented by special hardware including application-specific integrated circuits, special CPUs, special memories, special components, etc. In general, all functions performed by computer programs can be easily implemented with corresponding hardware, and the specific hardware structures used to implement the same function can also be diverse, such as analog circuits, digital circuits or special circuits, etc. However, for the present application, software program implementation is a better implementation method in most cases. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer's floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk or optical disk, etc., and includes a number of instructions to enable a computer device (which can be a personal computer, training equipment, or network equipment, etc.) to execute the methods described in each embodiment of the present application.

In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware, or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product.

The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, training equipment or data center to another website, computer, training equipment or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can store or a data storage device such as a training device, data center, etc. that includes one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

Claims (19)

A data processing method, characterized in that the method comprises: Acquire features of the video, where the features include multiple feature maps distributed in a time dimension, and the dimensions of the feature maps include a channel dimension and a spatial dimension; Converting the plurality of feature maps from being distributed in the time dimension to being distributed in the channel dimension or the spatial dimension to obtain one or more first feature maps; A task processing result of the video is obtained according to the one or more first feature maps. The method according to claim 1 is characterized in that the first feature map does not include the time dimension. The method according to claim 1 or 2 is characterized in that the task processing result is a processing result of a video understanding task, a video-based generation task, or a video enhancement task. The method according to any one of claims 1 to 3 is characterized in that the first feature map and the feature have the same size in the spatial dimension. The method according to any one of claims 1 to 4, characterized in that the size of the feature is (x, t, h, w), and the size of the first feature map is (x*t, h, w), wherein x is the number of channels, t is the number of times, h and w are the height and width in the spatial dimension, respectively, and * is the product. The method according to any one of claims 1 to 5, characterized in that converting the multiple feature maps from being distributed in the time dimension to being distributed in the channel dimension or the spatial dimension to obtain the one or more first feature maps includes: The feature maps at different time dimensions in the multiple feature maps are stacked in the channel dimension to obtain the one or more first feature maps. The method according to any one of claims 1 to 6, wherein obtaining the task processing result of the video based on the one or more first feature maps comprises: Processing the first feature map through a feature extraction network to obtain one or more second feature maps; According to the one or more second feature maps, a task processing result of the video is obtained through a task network. The method according to claim 7, wherein processing the first feature map through a feature extraction network to obtain one or more second feature maps comprises: Determining, based on an input feature map, a weight corresponding to a channel dimension of the input feature map, wherein the input feature map is an intermediate output obtained by the feature extraction network processing the one or more first feature maps; A convolution operation is performed on the input feature map to obtain convolution operation results of multiple channel dimensions, and the convolution operation results of the multiple channel dimensions are fused according to the weights to obtain a processing result. The method according to claim 7 or 8, wherein processing the first feature map through a feature extraction network to obtain one or more second feature maps comprises: Expanding an input feature map into multiple feature maps distributed in a time dimension, wherein the input feature map is an intermediate output obtained by the feature extraction network processing the one or more first feature maps; Feature interaction is performed between the multiple feature maps distributed in the obtained time dimension. The method according to any one of claims 1 to 9, wherein the interaction comprises: Interaction based on attention mechanism or interaction achieved through large kernel convolution. The method according to any one of claims 7 to 10, wherein processing the first feature map through a feature extraction network to obtain one or more second feature maps comprises: Determining a temporal encoding consistent with a size of the plurality of feature maps; Fusing the time code and the multiple feature maps to obtain multiple fused feature maps in the time dimension; Feature interaction is performed between the multiple feature maps in the fused time dimension. The method according to any one of claims 9 to 11 is characterized in that the interaction module is further used to fuse the interaction result obtained through the interaction with the input features. A data processing device, characterized in that the device comprises: a compression module, configured to obtain features of the video, the features comprising a plurality of feature maps distributed in a time dimension, the dimensions of the feature maps comprising a channel dimension and a spatial dimension, and convert the plurality of feature maps from being distributed in the time dimension to being distributed in the channel dimension or the spatial dimension, to obtain the one or more first feature maps; A processing module is used to obtain a task processing result of the video based on the one or more first feature maps. The device according to claim 13 is characterized in that the size of the feature is (x, t, h, w), and the size of the first feature map is (x*t, h, w), where x is the number of channels, t is the number of times, h and w are the height and width in the spatial dimension, respectively, and * is the product. The device according to claim 13 or 14, characterized in that the compression module is specifically used to: The feature maps at different time dimensions in the multiple feature maps are stacked in the channel dimension to obtain the one or more first feature maps. A computer storage medium, characterized in that the computer storage medium stores one or more instructions, which, when executed by one or more computers, enable the one or more computers to perform the operations of the method described in any one of claims 1 to 12. A computer program product, characterized in that it comprises computer-readable instructions, and when the computer-readable instructions are executed on a computer device, the computer device is caused to execute the method according to any one of claims 1 to 12. A system comprises at least one processor and at least one memory; the processor and the memory are connected via a communication bus and communicate with each other; The at least one memory is used to store code; The at least one processor is configured to execute the code to perform the method according to any one of claims 1 to 12. A chip comprising a processor, wherein the processor is used to support a data processing device to implement the method according to any one of claims 1 to 12.