WO2021004247A1 - 视频封面生成方法、装置及电子设备 - Google Patents

视频封面生成方法、装置及电子设备 Download PDF

Info

Publication number
WO2021004247A1
WO2021004247A1 PCT/CN2020/096719 CN2020096719W WO2021004247A1 WO 2021004247 A1 WO2021004247 A1 WO 2021004247A1 CN 2020096719 W CN2020096719 W CN 2020096719W WO 2021004247 A1 WO2021004247 A1 WO 2021004247A1
Authority
WO
WIPO (PCT)
Prior art keywords
images
target
image
video
cover
Prior art date
Application number
PCT/CN2020/096719
Other languages
English (en)
French (fr)
Inventor
黄凯
王长虎
Original Assignee
北京字节跳动网络技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 北京字节跳动网络技术有限公司 filed Critical 北京字节跳动网络技术有限公司
Publication of WO2021004247A1 publication Critical patent/WO2021004247A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
    • H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
    • H04N21/44—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream or rendering scenes according to encoded video stream scene graphs
    • H04N21/44008—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream or rendering scenes according to encoded video stream scene graphs involving operations for analysing video streams, e.g. detecting features or characteristics in the video stream
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
    • H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
    • H04N21/454—Content or additional data filtering, e.g. blocking advertisements
    • H04N21/4545—Input to filtering algorithms, e.g. filtering a region of the image
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
    • H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
    • H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
    • H04N21/4662—Learning process for intelligent management, e.g. learning user preferences for recommending movies characterized by learning algorithms
    • H04N21/4665—Learning process for intelligent management, e.g. learning user preferences for recommending movies characterized by learning algorithms involving classification methods, e.g. Decision trees
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
    • H04N21/47—End-user applications
    • H04N21/475—End-user interface for inputting end-user data, e.g. personal identification number [PIN], preference data
    • H04N21/4756—End-user interface for inputting end-user data, e.g. personal identification number [PIN], preference data for rating content, e.g. scoring a recommended movie

Definitions

  • the present disclosure relates to the field of image processing technology, and in particular to a method, device and electronic device for generating a video cover.
  • Image processing also known as image processing, is a technology that uses computers to perform images to achieve the desired results. Originated in the 1920s, generally digital image processing.
  • the main content of image processing technology includes image compression, enhanced restoration, and matching description recognition.
  • Common processing includes image digitization, image coding, image enhancement, image restoration, image segmentation, and image analysis.
  • Image processing is the act of using computers to process image information to meet people's visual psychology or application needs. It is widely used and is mostly used in surveying and mapping, atmospheric sciences, astronomy, beautiful pictures, and improving image recognition.
  • An application scenario of image processing is to select a video frame in a video as the video cover of the video. How to select a representative and high-quality video frame from a large number of video frames as the video cover becomes a need Technical problems solved.
  • the embodiments of the present disclosure provide a method, device and electronic device for generating a video cover, which at least partially solve the problems existing in the prior art.
  • embodiments of the present disclosure provide a method for generating a video cover, including:
  • sight line detection and open eye detection are performed on the target object contained in the multiple candidate cover images to filter out candidates that do not meet preset requirements Cover image
  • the target image is selected from the remaining candidate cover images after filtering as the final cover image of the target video.
  • the parsing of the target video to obtain multiple parsed images includes:
  • the video frame containing the target object is set as the analytic image.
  • the performing clustering processing on the multiple analytic images to obtain multiple candidate cover images includes:
  • each cluster an image that meets a preset condition is selected as the candidate cover image.
  • the performing k-type clustering calculation on the multiple analytic images includes:
  • the method before said performing line-of-sight detection and eye-opening detection on the target objects contained in the multiple candidate cover images, the method further includes:
  • the quality score of each of the plurality of candidate cover images is determined.
  • the performing line-of-sight detection and eye-opening detection on the target objects contained in the plurality of candidate cover images includes:
  • sight line detection and eye opening detection are performed on the target object.
  • the performing line of sight detection and eye opening detection on the target object based on the result of the modeling operation includes:
  • the performing line of sight detection and eye opening detection on the target object based on the result of the modeling operation includes:
  • the image quality-based evaluation network scores the quality of the remaining candidate cover images after filtering, and selects a target image from the remaining candidate cover images after filtering as the final target video Cover image, including:
  • the cover image candidate with the highest quality score is used as the final cover image of the target video.
  • a video cover generation device including:
  • the analysis module is used to analyze the target video to obtain multiple analysis images
  • a clustering module configured to perform clustering processing on the multiple analytic images to obtain multiple candidate cover images
  • the filtering module is configured to perform line-of-sight detection and eye-opening detection on the target objects contained in the multiple candidate cover images based on the key point detection results performed on the target objects contained in the candidate cover images, so as to filter out non-conformities Candidate cover images required by default;
  • the selection module is used for scoring the quality of the remaining candidate cover images after filtering based on the image quality evaluation network, and selecting the target image from the remaining candidate cover images after filtering as the final cover image of the target video.
  • an embodiment of the present disclosure also provides an electronic device, which includes:
  • At least one processor and,
  • a memory communicatively connected with the at least one processor; wherein,
  • the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute any of the foregoing first aspect or any implementation of the first aspect Video cover generation method.
  • embodiments of the present disclosure also provide a non-transitory computer-readable storage medium that stores computer instructions that are used to make the computer execute the first aspect or the first aspect described above.
  • a video cover generation method in any implementation manner of one aspect.
  • the embodiments of the present disclosure also provide a computer program product.
  • the computer program product includes a computing program stored on a non-transitory computer-readable storage medium.
  • the computer program includes program instructions. When executing, the computer is caused to execute the video cover generation method in the foregoing first aspect or any implementation manner of the first aspect.
  • the video cover generation solution in the embodiment of the present disclosure includes analyzing the target video to obtain multiple analytical images; performing clustering processing on the multiple analytical images to obtain multiple candidate cover images; The result of the key point detection performed on the target object included in the multiple candidate cover images is performed on the target objects included in the plurality of candidate cover images to perform sight line detection and open eye detection to filter out candidate cover images that do not meet the preset requirements; based on image quality evaluation
  • the network scores the quality of the remaining candidate cover images after filtering, and selects the target image from the remaining candidate cover images after filtering as the final cover image of the target video.
  • FIG. 1 is a schematic diagram of a process for generating a video cover provided by an embodiment of the disclosure
  • FIG. 2 is a schematic diagram of another video cover generation process provided by an embodiment of the disclosure.
  • FIG. 3 is a schematic diagram of another video cover generation process provided by an embodiment of the disclosure.
  • FIG. 4 is a schematic diagram of another video cover generation process provided by an embodiment of the disclosure.
  • FIG. 5 is a schematic structural diagram of a video cover generation device provided by an embodiment of the disclosure.
  • FIG. 6 is a schematic diagram of an electronic device provided by an embodiment of the disclosure.
  • the embodiment of the present disclosure provides a method for generating a video cover.
  • the video cover generation method provided in this embodiment can be executed by a computing device, which can be implemented as software, or a combination of software and hardware, and the computing device can be integrated in a server, terminal device, etc.
  • a method for generating a video cover provided by an embodiment of the present disclosure includes:
  • the target video is a video file with audio and image data.
  • the target video can be a video file in any format.
  • the target video can be a video file in mpg, mp4, rm, rmvb, wax format, or other formats.
  • Video files are a video file with audio and image data.
  • the target video contains video frames.
  • the video frame is a collection of all the video pictures in the target video. For example, for a video with a frame rate of 30fps, according to the normal video playback speed, a 1-second video can be split 30 video frames out. Of course, based on actual needs, it is also possible to obtain more video frames by inserting frames in 30 video frames, or select some video frames from the 30 video frames.
  • the video frames in the target video may be filtered, and the video frames that meet the filtering conditions are used as the analysis image.
  • target detection can be performed on all video frames included in the target video. Through target detection, it can be determined whether the video frame contains a target object (for example, a person). When the video frame contains a target object, it will include The video frame of the target object is set as the analytic image. In this way, the pertinence of multiple analytical images can be further improved.
  • a variety of image-specific target detection methods can be used to perform target detection on video frames, and the method of target detection is not limited here.
  • S102 Perform clustering processing on the multiple analysis images to obtain multiple candidate cover images.
  • the analytical images all contain target objects.
  • clustering can be performed on multiple analytical images. By clustering the images, a part of the typical analysis images can be selected The image (multiple candidate cover images) is further processed.
  • k categories may be set for a plurality of analysis images in advance to perform clustering calculation.
  • key point detection can be performed on the target object in the candidate cover image.
  • the key points of the head of the target image can be obtained.
  • the target object can be detected The line of sight in the candidate cover image and whether the eyes are open are detected, thereby filtering out the images with the target object's line of sight shifted or eyes not fully opened.
  • the key point detection results performed on the target object contained in the candidate cover image can be obtained, and based on the key point detection results, the pupil key points, the eyeball key points, and the eye contour key points can be determined; based on the pupil key points, The eyeball key points and the eye contour key points perform a modeling operation on the eyeballs of the target object, and through the result of the modeling operation, the sight line detection and the eye opening detection can be performed on the target object.
  • the CLNF (constrained local neural fields) model can be used to detect the key points of the eyeball through the results of the modeling operation, and at the same time perform 3D modeling of the eye, and after completing the 3D modeling of the eye, connect the origin to the pupil
  • the center forms a ray, calculate the intersection point with the eyeball, and use the vector from the center of the eyeball to the intersection point as the direction of the line of sight. In order to further determine whether the target object's line of sight meets the requirements.
  • the aspect ratio of the outer eye contour may be calculated based on the result of the modeling operation, and whether the target object is in an open eye state can be determined by determining whether the aspect ratio is greater than a preset threshold.
  • the candidate cover images that do not meet the requirements can be filtered out, so as to obtain the set of candidate cover images remaining after filtering.
  • S104 Based on the quality score of the remaining candidate cover images after the filtering by the image quality evaluation network, a target image is selected from the remaining candidate cover images after the filtering as the final cover image of the target video.
  • an image can be further filtered from the set to determine the final cover image of the target video.
  • a pre-trained convolutional neural network is used to score all the candidate cover images.
  • the process of quality scoring Can comprehensively evaluate the image quality, color, environment and expression, etc., so as to give a specific quality score for each candidate cover image.
  • one or more candidate cover images with high quality scores can be selected as the final cover image of the target video by sorting.
  • the video frames in the target video can be filtered in a variety of ways, so as to filter the final cover image that matches the target video.
  • parsing the target video to obtain multiple parsed images may include the following steps:
  • S201 Perform target detection on all video frames included in the target video.
  • Target detection can be performed on all the video frames contained in the target video. Through target detection, it can be determined whether the video frame contains a target object (for example, a person). Target detection on the video frame can use a variety of image-specific target detection methods. , The method of target detection is not limited here.
  • S202 Based on the result of the target detection, determine whether the target object is included in the video frames constituting the target video.
  • multiple object detection results can be obtained, and by comparing the multiple object detection results with the target object for similarity, it is possible to further determine whether the target object is contained in the video frame of the target video.
  • the video frame images can be selected in a targeted manner to obtain multiple analytical images.
  • k-type clustering calculations may be performed on the multiple analytic images, and in each cluster Select an image that meets the preset conditions as the candidate cover image.
  • performing k-type clustering calculation on the multiple analytic images may include the following steps:
  • S302 Use the nearest neighbor rule to allocate all samples to k classes ⁇ j(k) represented by each cluster center, and the number of samples contained in each class is Nj(l);
  • the method before the line of sight detection and eye opening detection are performed on the target objects contained in the plurality of candidate cover images, the method further includes: using a preset convolutional nerve The network performs quality evaluation on the multiple candidate cover images; based on the result of the quality evaluation, the quality score of each of the multiple candidate cover images is determined.
  • performing line-of-sight detection and eye-opening detection on the target objects included in the multiple candidate cover images includes:
  • S401 Acquire a key point detection result performed on a target object included in a candidate cover image.
  • key point detection can be performed on the head region of the target object in the cover image.
  • key point detection key point data of multiple organs (for example, eyes) in the head region can be obtained.
  • S402 Determine the pupil key point, the eyeball key point, and the eye contour key point based on the key point detection result.
  • the key points of the exit pupil, the key points of the eyeball and the key points of the eye contour can be determined.
  • S403 Perform a modeling operation on the eyeball of the target object based on the pupil key points, the eyeball key points, and the eye contour key points.
  • the eyeball of the target object can be modeled based on the key point data. After modeling, the eyeball of the target object can be quantified in more detail.
  • the modeling operation of the target object's eyeballs can be performed in various ways, which are not limited here.
  • S404 Perform line-of-sight detection and eye-opening detection on the target object based on the result of the modeling operation.
  • the CLNF model can be used to detect the key points of the eyeball through the results of the modeling operation, and the eye can be three-dimensionally modeled through the key points of the eyeball.
  • the model after the three-dimensional modeling connects the origin to the pupil center A ray is formed, and the intersection point with the eyeball is calculated, and the vector from the center of the eyeball to the direction of the intersection point is taken as the line of sight direction. Therefore, based on the determined direction of the line of sight, it is determined whether the line of sight of the target object meets the line of sight requirement.
  • the aspect ratio of the outer eye contour can also be calculated based on the result of the modeling operation, and whether the target object is in an open eye state can be determined by determining whether the aspect ratio is greater than a preset threshold. As a result, candidate cover images that do not meet the open eye state are deleted.
  • the image quality evaluation network scores the quality of the remaining candidate cover images after filtering, and selects a target image from the remaining candidate cover images after filtering as the target video
  • the final cover image includes: selecting the candidate cover image with the highest quality score from the remaining candidate cover images after filtering; and using the candidate cover image with the highest quality score as the final cover image of the target video.
  • an embodiment of the present disclosure also provides a video cover generation device 50, including:
  • the analysis module 501 is used to analyze the target video to obtain multiple analysis images.
  • the target video is a video file with audio and image data.
  • the target video can be a video file in any format.
  • the target video can be a video file in mpg, mp4, rm, rmvb, wax format, or other formats.
  • Video files are a video file with audio and image data.
  • the target video contains video frames.
  • the video frame is a collection of all the video pictures in the target video. For example, for a video with a frame rate of 30fps, according to the normal video playback speed, a 1-second video can be split 30 video frames out. Of course, based on actual needs, it is also possible to obtain more video frames by inserting frames in 30 video frames, or select some video frames from the 30 video frames.
  • the video frames in the target video may be filtered, and the video frames that meet the filtering conditions are used as the analysis image.
  • target detection can be performed on all video frames included in the target video. Through target detection, it can be determined whether the video frame contains a target object (for example, a person). When the video frame contains a target object, it will include The video frame of the target object is set as the analytic image. In this way, the pertinence of multiple analytical images can be further improved.
  • a variety of image-specific target detection methods can be used to perform target detection on video frames, and the method of target detection is not limited here.
  • the clustering module 502 is configured to perform clustering processing on the multiple analytic images to obtain multiple candidate cover images.
  • the analytical images all contain target objects.
  • clustering can be performed on multiple analytical images. By clustering the images, a part of the typical analysis images can be selected The image (multiple candidate cover images) is further processed.
  • k categories may be set for a plurality of analysis images in advance to perform clustering calculation.
  • the filtering module 503 is configured to perform line-of-sight detection and eye-opening detection on the target objects included in the multiple candidate cover images based on the key point detection results performed on the target objects included in the candidate cover image, so as to filter out Candidate cover images that meet the preset requirements.
  • key point detection can be performed on the target object in the candidate cover image.
  • the key points of the head of the target image can be obtained.
  • the target object can be detected The line of sight in the candidate cover image and whether the eyes are open are detected, thereby filtering out the images with the target object's line of sight shifted or eyes not fully opened.
  • the key point detection results performed on the target object contained in the candidate cover image can be obtained, and based on the key point detection results, the pupil key points, the eyeball key points, and the eye contour key points can be determined; based on the pupil key points, The eyeball key points and the eye contour key points perform a modeling operation on the eyeballs of the target object, and through the result of the modeling operation, the sight line detection and the eye opening detection can be performed on the target object.
  • the CLNF model can be used to detect the key points of the eyeball through the results of the modeling operation, and the eye can be 3D modeled, and after the 3D modeling of the eye is completed, connect the origin to the pupil center A ray is formed, and the intersection point with the eyeball is calculated, and the vector from the center of the eyeball to the direction of the intersection point is taken as the line of sight direction. In order to further determine whether the target object's line of sight meets the requirements.
  • the aspect ratio of the outer eye contour may be calculated based on the result of the modeling operation, and whether the target object is in an open eye state can be determined by determining whether the aspect ratio is greater than a preset threshold.
  • the candidate cover images that do not meet the requirements can be filtered out, so as to obtain the set of candidate cover images remaining after filtering.
  • the selection module 504 is configured to score the quality of the remaining candidate cover images after filtering based on the image quality evaluation network, and select the target image from the remaining candidate cover images after filtering as the final cover image of the target video.
  • an image can be further filtered from the set to determine the final cover image of the target video.
  • a pre-trained convolutional neural network is used to score all the candidate cover images.
  • the process of quality scoring Can comprehensively evaluate the image quality, color, environment and expression, etc., so as to give a specific quality score for each candidate cover image.
  • one or more candidate cover images with high quality scores can be selected as the final cover image of the target video by sorting.
  • the device shown in FIG. 5 can correspondingly execute the content in the foregoing method embodiment.
  • an electronic device 60 which includes:
  • At least one processor and,
  • a memory communicatively connected with the at least one processor; wherein,
  • the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute the video cover generation method in the foregoing method embodiment.
  • the embodiments of the present disclosure also provide a non-transitory computer-readable storage medium that stores computer instructions, and the computer instructions are used to make the computer execute the foregoing method embodiments.
  • the embodiments of the present disclosure also provide a computer program product, the computer program product includes a calculation program stored on a non-transitory computer-readable storage medium, the computer program includes program instructions, when the program instructions are executed by a computer, The computer executes the video cover generation method in the foregoing method embodiment.
  • FIG. 6 shows a schematic structural diagram of an electronic device 60 suitable for implementing embodiments of the present disclosure.
  • Electronic devices in the embodiments of the present disclosure may include, but are not limited to, mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablets), PMPs (portable multimedia players), vehicle-mounted terminals (for example, Mobile terminals such as car navigation terminals) and fixed terminals such as digital TVs, desktop computers, etc.
  • the electronic device shown in FIG. 6 is only an example, and should not bring any limitation to the function and scope of use of the embodiments of the present disclosure.
  • the electronic device 60 may include a processing device (such as a central processing unit, a graphics processor, etc.) 601, which may be loaded into a random access device according to a program stored in a read-only memory (ROM) 602 or from a storage device 608.
  • the program in the memory (RAM) 603 executes various appropriate actions and processing.
  • the RAM 603 also stores various programs and data required for the operation of the electronic device 60.
  • the processing device 601, the ROM 602, and the RAM 603 are connected to each other through a bus 604.
  • An input/output (I/O) interface 605 is also connected to the bus 604.
  • the following devices can be connected to the I/O interface 605: including input devices 606 such as touch screen, touch panel, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; including, for example, liquid crystal display (LCD), speakers, An output device 607 such as a vibrator; a storage device 608 such as a magnetic tape, a hard disk, etc.; and a communication device 609.
  • the communication device 609 may allow the electronic device 60 to perform wireless or wired communication with other devices to exchange data.
  • the figure shows the electronic device 60 with various devices, it should be understood that it is not required to implement or have all the devices shown. It may alternatively be implemented or provided with more or fewer devices.
  • the process described above with reference to the flowchart can be implemented as a computer software program.
  • the embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart.
  • the computer program may be downloaded and installed from the network through the communication device 609, or installed from the storage device 608, or installed from the ROM 602.
  • the computer program is executed by the processing device 601, it executes the above-mentioned functions defined in the method of the embodiment of the present disclosure.
  • the above-mentioned computer-readable medium in the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two.
  • the computer-readable storage medium may be, for example, but not limited to, an electric, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable Programmable read only memory (EPROM or flash memory), optical fiber, portable compact disk read only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.
  • a computer-readable storage medium may be any tangible medium that contains or stores a program, and the program may be used by or in combination with an instruction execution system, apparatus, or device.
  • a computer-readable signal medium may include a data signal propagated in a baseband or as a part of a carrier wave, and a computer-readable program code is carried therein. This propagated data signal can take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing.
  • the computer-readable signal medium may also be any computer-readable medium other than the computer-readable storage medium.
  • the computer-readable signal medium may send, propagate, or transmit the program for use by or in combination with the instruction execution system, apparatus, or device .
  • the program code contained on the computer-readable medium can be transmitted by any suitable medium, including but not limited to: wire, optical cable, RF (Radio Frequency), etc., or any suitable combination of the above.
  • the above-mentioned computer-readable medium may be included in the above-mentioned electronic device; or it may exist alone without being assembled into the electronic device.
  • the above-mentioned computer-readable medium carries one or more programs, and when the above-mentioned one or more programs are executed by the electronic device, the electronic device: obtains at least two Internet protocol addresses; and sends to the node evaluation device including the at least two A node evaluation request for an Internet Protocol address, wherein the node evaluation device selects an Internet Protocol address from the at least two Internet Protocol addresses and returns it; receives the Internet Protocol address returned by the node evaluation device; wherein, the obtained The Internet Protocol address indicates the edge node in the content distribution network.
  • the aforementioned computer-readable medium carries one or more programs, and when the aforementioned one or more programs are executed by the electronic device, the electronic device: receives a node evaluation request including at least two Internet Protocol addresses; Among the at least two Internet Protocol addresses, select an Internet Protocol address; return the selected Internet Protocol address; wherein, the received Internet Protocol address indicates an edge node in the content distribution network.
  • the computer program code used to perform the operations of the present disclosure may be written in one or more programming languages or a combination thereof.
  • the above-mentioned programming languages include object-oriented programming languages—such as Java, Smalltalk, C++, and also conventional Procedural programming language-such as "C" language or similar programming language.
  • the program code can be executed entirely on the user's computer, partly on the user's computer, executed as an independent software package, partly on the user's computer and partly executed on a remote computer, or entirely executed on the remote computer or server.
  • the remote computer can be connected to the user’s computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (for example, using an Internet service provider to pass Internet connection).
  • LAN local area network
  • WAN wide area network
  • each block in the flowchart or block diagram can represent a module, program segment, or part of code, and the module, program segment, or part of code contains one or more for realizing the specified logical function Executable instructions.
  • the functions marked in the block may also occur in a different order from the order marked in the drawings. For example, two blocks shown in succession can actually be executed substantially in parallel, or they can sometimes be executed in the reverse order, depending on the functions involved.
  • each block in the block diagram and/or flowchart, and the combination of the blocks in the block diagram and/or flowchart can be implemented by a dedicated hardware-based system that performs the specified functions or operations Or it can be realized by a combination of dedicated hardware and computer instructions.
  • the units involved in the embodiments described in the present disclosure may be implemented in a software manner, or may be implemented in a hardware manner.
  • the name of the unit does not constitute a limitation on the unit itself under certain circumstances.
  • the first obtaining unit can also be described as "a unit for obtaining at least two Internet Protocol addresses.”

Landscapes

  • Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Human Computer Interaction (AREA)
  • Image Analysis (AREA)

Abstract

本公开实施例中提供了一种视频封面生成方法、装置及电子设备，属于图像处理技术领域，该方法包括：对目标视频进行解析，得到多个解析图像；对所述多个解析图像进行聚类处理，得到多个候选封面图像；基于对所述候选封面图像中包含的目标对象执行的关键点检测结果，对所述多个候选封面图像中包含的目标对象执行视线检测和睁眼检测，以过滤掉不符合预设要求的候选封面图像；基于图像质量评价网络对过滤后剩余的候选封面图像的质量评分，从过滤后剩余的候选封面图像中选取目标图像作为所述目标视频的最终封面图像。通过本公开的处理方案，能够自动生成高质量的视频封面。

Description

视频封面生成方法、装置及电子设备
相关申请的交叉引用
本申请要求于2019年07月11日提交的，申请号为201910622565.6、发明名称为“视频封面生成方法、装置及电子设备”的中国专利申请的优先权，该申请的全文通过引用结合在本申请中。
技术领域
本公开涉及图像处理技术领域，尤其涉及一种视频封面生成方法、装置及电子设备。
背景技术
图像处理(image processing)又称为影像处理，是用计算机对图像进行达到所需结果的技术。起源于20世纪20年代，一般为数字图像处理。图像处理技术的主要内容包括图像压缩、增强复原、匹配描述识别3个部分，常见的处理有图像数字化、图像编码、图像增强、图像复原、图像分割和图像分析等。图像处理是利用计算机对图像信息进行加工以满足人的视觉心理或者应用需求的行为，应用广泛，多用于测绘学、大气科学、天文学、美图、使图像提高辨识等。
图像处理的一个应用场景便是在一段视频中选择一个视频帧作为该段视频的视频封面，如何能够从众多的视频帧中选择一个具有代表性、图像质量高的视频帧作为视频封面，成为需要解决的技术问题。
发明内容
有鉴于此，本公开实施例提供一种视频封面生成方法、装置及电子设备，至少部分解决现有技术中存在的问题。
第一方面，本公开实施例提供了一种视频封面生成方法，包括：
对目标视频进行解析，得到多个解析图像；
对所述多个解析图像进行聚类处理，得到多个候选封面图像；
基于对所述候选封面图像中包含的目标对象执行的关键点检测结果，对所述多个候选封面图像中包含的目标对象执行视线检测和睁眼检测，以过滤掉不符合预设要求的候选封面图像；
基于图像质量评价网络对过滤后剩余的候选封面图像的质量评分，从过滤后剩余的候选封面图像中选取目标图像作为所述目标视频的最终封面图像。
根据本公开实施例的一种具体实现方式，所述对目标视频进行解析，得到多个解析图像，包括：
对所述目标视频中包含的所有视频帧执行目标检测；
基于目标检测的结果，判断组成目标视频的视频帧中是否包含所述目标对象；
若是，则将包含所述目标对象的视频帧设置为所述解析图像。
根据本公开实施例的一种具体实现方式，所述对所述多个解析图像进行聚类处理，得到多个候选封面图像，包括：
对所述多个解析图像执行k类的聚类计算；
在每个聚类中选择一张符合预设条件的图像作为所述候选封面图像。
根据本公开实施例的一种具体实现方式，所述对所述多个解析图像执行k类的聚类计算，包括：
在所述多个解析图像中选取k个样本点为初始聚类中心，记为z1(l),z2(l),……zk(l)，迭代序号l＝1；
使用最近邻规则将所有样本分配到各聚类中心所代表的k类ωj(k)中，各类所包含的样本数为Nj(l)；
计算各类的重心，将计算得到的重心确定为新的聚类中心；
对于第j次迭代，判断zj(l+1)与zj(l)的值是否相同，当zj(l+1)≠zj(l)时继续迭代计算，当zj(l+1)＝zj(l)时，停止迭代计算。
根据本公开实施例的一种具体实现方式，所述对所述多个候选封面图像中包含的目标对象执行视线检测和睁眼检测之前，所述方法还包括：
利用预先设置的卷积神经网络对所述多个候选封面图像执行质量评价；
基于所述质量评价的结果，确定所述多个候选封面图像中每一个图像的质量评分。
根据本公开实施例的一种具体实现方式，所述对所述多个候选封面图像中包含的目标对象执行视线检测和睁眼检测，包括：
获取对候选封面图像中包含的目标对象执行的关键点检测结果；
基于所述关键点检测结果，确定出瞳孔关键点、眼球关键点及眼轮廓关键点；
基于所述瞳孔关键点、眼球关键点及眼轮廓关键点，对所述目标对象的眼球执行建模操作；
基于建模操作的结果，来对所述目标对象执行视线检测和睁眼检测。
根据本公开实施例的一种具体实现方式，所述基于建模操作的结果，来对所述目标对象执行视线检测和睁眼检测，包括：
使用CLNF模型检测出眼球的关键点；
对眼睛进行三维建模；
连接原点到瞳孔中心形成一条射线，计算其与眼球的交点，将眼球中心到交点方向的向量作为视线方向。
根据本公开实施例的一种具体实现方式，所述基于建模操作的结果，来对所述目标对象执行视线检测和睁眼检测，包括：
通过建模操作的结果，计算外眼轮廓的宽高比；
通过判断所述宽高比是否大于预设阈值来判断所述目标对象是否处于睁眼状态。
根据本公开实施例的一种具体实现方式，所述基于图像质量评价网络对过滤后剩余的候选封面图像的质量评分，从过滤后剩余的候选封面图像中选取目标图像作为所述目标视频的最终封面图像，包括：
从过滤后剩余的候选封面图像中选取质量评分最高的候选封面图像；
将所述质量评分最高的候选封面图像作为所述目标视频的最终封面图像。
第二方面，本公开实施例提供了一种视频封面生成装置，包括：
解析模块，用于对目标视频进行解析，得到多个解析图像；
聚类模块，用于对所述多个解析图像进行聚类处理，得到多个候选封面图像；
过滤模块，用于基于对所述候选封面图像中包含的目标对象执行的关键点检测结果，对所述多个候选封面图像中包含的目标对象执行视线检测和睁眼检测，以过滤掉不符合预设要求的候选封面图像；
选择模块，用于基于图像质量评价网络对过滤后剩余的候选封面图像的质量评分，从过滤后剩余的候选封面图像中选取目标图像作为所述目标视频的最终封面图像。
第三方面，本公开实施例还提供了一种电子设备，该电子设备包括：
至少一个处理器；以及，
与该至少一个处理器通信连接的存储器；其中，
该存储器存储有可被该至少一个处理器执行的指令，该指令被该至少一个处理器执行，以使该至少一个处理器能够执行前述任第一方面或第一方面的任一实现方式中的视频封面生成方法。
第四方面，本公开实施例还提供了一种非暂态计算机可读存储介质，该非暂态计算机可读存储介质存储计算机指令，该计算机指令用于使该计算机执行前述第一方面或第一方面的任一实现方式中的视频封面生成方法。
第五方面，本公开实施例还提供了一种计算机程序产品，该计算机程序产品包括存储在非暂态计算机可读存储介质上的计算程序，该计算机程序包括程序指令，当该程序指令被计算机执行时，使该计算机执行前述第一方面或第一方面的任一实现方式中的视频封面生成方法。
本公开实施例中的视频封面生成方案，包括对目标视频进行解析，得到多个解析图像；对所述多个解析图像进行聚类处理，得到多个候选封面图像；基于对所述候选封面图像中包含的目标对象执行的关键点检测结果，对所述多个候选封面图像中包含的目标对象执行视线检测和睁眼检测，以过滤掉不符合预 设要求的候选封面图像；基于图像质量评价网络对过滤后剩余的候选封面图像的质量评分，从过滤后剩余的候选封面图像中选取目标图像作为所述目标视频的最终封面图像。通过本公开的方案，能够自动的选择高质量的典型视频帧作为视频的封面图像。
附图说明
为了更清楚地说明本公开实施例的技术方案，下面将对实施例中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本公开的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其它的附图。
图1为本公开实施例提供的一种视频封面生成流程示意图；
图2为本公开实施例提供的另一种视频封面生成流程示意图；
图3为本公开实施例提供的另一种视频封面生成流程示意图；
图4为本公开实施例提供的另一种视频封面生成流程示意图；
图5为本公开实施例提供的一种视频封面生成装置结构示意图；
图6为本公开实施例提供的电子设备示意图。
具体实施方式
下面结合附图对本公开实施例进行详细描述。
以下通过特定的具体实例说明本公开的实施方式，本领域技术人员可由本说明书所揭露的内容轻易地了解本公开的其他优点与功效。显然，所描述的实施例仅仅是本公开一部分实施例，而不是全部的实施例。本公开还可以通过另外不同的具体实施方式加以实施或应用，本说明书中的各项细节也可以基于不同观点与应用，在没有背离本公开的精神下进行各种修饰或改变。需说明的是，在不冲突的情况下，以下实施例及实施例中的特征可以相互组合。基于本公开中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本公开保护的范围。
需要说明的是，下文描述在所附权利要求书的范围内的实施例的各种方面。应显而易见，本文中所描述的方面可体现于广泛多种形式中，且本文中所描述 的任何特定结构及/或功能仅为说明性的。基于本公开，所属领域的技术人员应了解，本文中所描述的一个方面可与任何其它方面独立地实施，且可以各种方式组合这些方面中的两者或两者以上。举例来说，可使用本文中所阐述的任何数目个方面来实施设备及/或实践方法。另外，可使用除了本文中所阐述的方面中的一或多者之外的其它结构及/或功能性实施此设备及/或实践此方法。
还需要说明的是，以下实施例中所提供的图示仅以示意方式说明本公开的基本构想，图式中仅显示与本公开中有关的组件而非按照实际实施时的组件数目、形状及尺寸绘制，其实际实施时各组件的型态、数量及比例可为一种随意的改变，且其组件布局型态也可能更为复杂。
另外，在以下描述中，提供具体细节是为了便于透彻理解实例。然而，所属领域的技术人员将理解，可在没有这些特定细节的情况下实践所述方面。
本公开实施例提供一种视频封面生成方法。本实施例提供的视频封面生成方法可以由一计算装置来执行，该计算装置可以实现为软件，或者实现为软件和硬件的组合，该计算装置可以集成设置在服务器、终端设备等中。
参见图1，本公开实施例提供的一种视频封面生成方法，包括：
S101，对目标视频进行解析，得到多个解析图像。
目标视频是一段记载有音频和影像资料的视频文件，目标视频可以是任意格式的视频文件，例如，目标视频可以是mpg、mp4、rm、rmvb、wax格式的视频文件，也可以是其他格式的视频文件。
目标视频中包含有视频帧，视频帧是目标视频中所有视频图片的集合，例如，对于一个帧率为30fps的视频而言，按照正常的视频播放速度，1秒钟长度的视频中可以拆分出30个视频帧。当然基于实际的需要，也可以通过在30个视频帧中进行插帧的方式，获得更多的视频帧，或者，从30个视频帧中选择部分视频帧。
在对目标视频进行解析的过程中，为了提高解析的效率，可以对目标视频中的视频帧进行筛选，将符合筛选条件的视频帧作为解析图像。作为一个例子，可以对所述目标视频中包含的所有视频帧执行目标检测，通过目标检测，可以 判断视频帧是否包含目标对象(例如，人)，当视频帧中包含目标对象时，则将包含所述目标对象的视频帧设置为所述解析图像。通过这种方式，能够进一步的提高多个解析图像的针对性。对视频帧进行目标检测可以采用多种针对图像的目标检测方法，在此对目标检测的方式不作限定。
S102，对所述多个解析图像进行聚类处理，得到多个候选封面图像。
解析图像中均含有目标对象，为了能够对包含目标对象的解析图像做进一步的筛选，可以对多个解析图像执行聚类处理，通过对图像进行聚类，可以在多个解析图像中选择一部分典型图像(多个候选封面图像)做进一步的处理。
具体的，可以预先对多个解析图像设置k类，从而进行聚类计算。首先，在所述多个解析图像中选取k个样本点为初始聚类中心，记为z1(l),z2(l),……zk(l)，迭代序号l＝1；接下来，使用最近邻规则将所有样本分配到各聚类中心所代表的k类ωj(k)中，各类所包含的样本数为Nj(l)；接下来，通过计算各类的重心的方式，将计算得到的重心确定为新的聚类中心；最后，通过循环迭代的方式对于第j次迭代，判断zj(l+1)与zj(l)的值是否相同，当zj(l+1)≠zj(l)时继续迭代计算，当zj(l+1)＝zj(l)时，停止迭代计算。在迭代停止之后，可以通过最终聚类的结果，在每个聚类中选择一个解析图像，最终形成k个候选封面图像。
S103，基于对所述候选封面图像中包含的目标对象执行的关键点检测结果，对所述多个候选封面图像中包含的目标对象执行视线检测和睁眼检测，以过滤掉不符合预设要求的候选封面图像。
获取到多个候选封面图像之后，可以对候选封面图像中的目标对象执行关键点检测，通过关键点检测，可以得到目标图像头部的关键点，基于头部的关键点，可以对目标对象在候选封面图像中的视线和是否睁眼进行检测，从而将目标对象视线偏移或眼睛没有完全睁开的图像过滤掉。
具体的，可以获取对候选封面图像中包含的目标对象执行的关键点检测结果，基于这些关键点检测结果，确定出瞳孔关键点、眼球关键点及眼轮廓关键点；基于所述瞳孔关键点、眼球关键点及眼轮廓关键点，对所述目标对象的眼 球执行建模操作，通过建模操作的结果，能够对所述目标对象执行视线检测和睁眼检测。
作为一种方式，可以通过建模操作的结果使用CLNF(constrained local neural fields)模型检测出眼球的关键点，同时对眼睛进行三维建模，并在对眼睛完成三维建模之后，连接原点到瞳孔中心形成一条射线，计算其与眼球的交点，将眼球中心到交点方向的向量作为视线方向。从而进一步的判断目标对象的视线是否符合要求。
作为另外一种方式，可以通过建模操作的结果，计算外眼轮廓的宽高比，通过判断所述宽高比是否大于预设阈值来判断所述目标对象是否处于睁眼状态。
当通过视线检测和睁眼检测发现候选封面图像不符合要求时，可以将不符合要求的候选封面图像过滤掉，从而得到过滤后剩余的候选封面图像集合。
S104，基于图像质量评价网络对过滤后剩余的候选封面图像的质量评分，从过滤后剩余的候选封面图像中选取目标图像作为所述目标视频的最终封面图像。
得到过滤后剩余的候选封面图像集合之后，可以从该集合中进一步的筛选一个图像，从而确定目标视频的最终封面图像。
作为一种方式，可以对所述多个候选封面图像中包含的目标对象执行视线检测和睁眼检测之前，通过使用预先训练的卷积神经网络对所有候选封面图像进行质量评分，质量评分的过程中，可以从图像画质、色彩、环境和表情等多个方面进行综合评价，从而对每个候选封面图像给出一个具体的质量评分。
在得到过滤后剩余的候选封面图像集合中每一个图像的质量评分之后，可以通过排序的方式，选择质量评分高的一个或多个候选封面图像作为该目标视频的最终封面图像。
通过上面的步骤，能够通过多种方式对目标视频中的视频帧进行过滤筛选，从而筛选出于目标视频相匹配的最终封面图像。
根据本公开实施例的一种可选实现方式，参见图2，对目标视频进行解析，得到多个解析图像，可以包括如下步骤：
S201，对所述目标视频中包含的所有视频帧执行目标检测。
可以对所述目标视频中包含的所有视频帧执行目标检测，通过目标检测，可以判断视频帧是否包含目标对象(例如，人)，对视频帧进行目标检测可以采用多种针对图像的目标检测方法，在此对目标检测的方式不作限定。
S202，基于目标检测的结果，判断组成目标视频的视频帧中是否包含所述目标对象。
通过对视频帧执行目标检测，可以获得多个对象检测结果，通过将多个对象检测结果与目标对象进行相似度比对，可以进一步的判断目标视频的视频帧中是否含有目标对象。
S203，若是，则将包含所述目标对象的视频帧设置为所述解析图像。
通过步骤S201-203中的步骤，能够有针对性的对视频帧图像进行选择，获得多个解析图像。
作为另外一种情况，在对所述多个解析图像进行聚类处理，得到多个候选封面图像的过程中，可以对所述多个解析图像执行k类的聚类计算，在每个聚类中选择一张符合预设条件的图像作为所述候选封面图像。
根据本公开实施例的一种可选实现方式，参见图3，对所述多个解析图像执行k类的聚类计算，可以包括如下步骤：
S301，在所述多个解析图像中选取k个样本点为初始聚类中心，记为z1(l),z2(l),……zk(l)，迭代序号l＝1；
S302，使用最近邻规则将所有样本分配到各聚类中心所代表的k类ωj(k)中，各类所包含的样本数为Nj(l)；
S303，计算各类的重心，将计算得到的重心确定为新的聚类中心；
S304，对于第j次迭代，判断zj(l+1)与zj(l)的值是否相同，当zj(l+1)≠zj(l)时继续迭代计算，当zj(l+1)＝zj(l)时，停止迭代计算。
根据本公开实施例的一种可选实现方式，所述对所述多个候选封面图像中包含的目标对象执行视线检测和睁眼检测之前，所述方法还包括：利用预先设置的卷积神经网络对所述多个候选封面图像执行质量评价；基于所述质量评价 的结果，确定所述多个候选封面图像中每一个图像的质量评分。
参见图4，根据本公开实施例的一种可选实现方式，对所述多个候选封面图像中包含的目标对象执行视线检测和睁眼检测，包括：
S401，获取对候选封面图像中包含的目标对象执行的关键点检测结果。
具体的，可以针对封面图像中目标对象的头部区域执行关键点检测，通过关键点检测，能够获得头部区域多个器官(例如，眼睛)的关键点数据。
S402，基于所述关键点检测结果，确定出瞳孔关键点、眼球关键点及眼轮廓关键点。
通过对检测到的关键点进行目标识别，能够确定出出瞳孔关键点、眼球关键点及眼轮廓关键点。
S403，基于所述瞳孔关键点、眼球关键点及眼轮廓关键点，对所述目标对象的眼球执行建模操作。
瞳孔关键点、眼球关键点及眼轮廓关键点之后，可以基于这些关键点数据对目标对象的眼球执行建模操作，建模之后，能够对目标对象的眼球进行更加细致的量化。对于目标对象眼球的建模操作可以采用多种方式进行，在此不作限定。
S404，基于建模操作的结果，来对所述目标对象执行视线检测和睁眼检测。
在实现步骤S404的过程中，可以通过建模操作的结果使用CLNF模型检测出眼球的关键点，通过眼球的关键点对眼睛进行三维建模，对三维建模之后的模型，连接原点到瞳孔中心形成一条射线，计算其与眼球的交点，将眼球中心到交点方向的向量作为视线方向。从而基于确定的视线的方向判断目标对象的视线是否满足视线要求。
除了进行视线检测之外，还可以通过建模操作的结果，计算外眼轮廓的宽高比，通过判断所述宽高比是否大于预设阈值来判断所述目标对象是否处于睁眼状态。从而删除掉不符合睁眼状态的候选封面图像。
根据本公开实施例的一种可选实现方式，所述基于图像质量评价网络对过滤后剩余的候选封面图像的质量评分，从过滤后剩余的候选封面图像中选取目 标图像作为所述目标视频的最终封面图像，包括：从过滤后剩余的候选封面图像中选取质量评分最高的候选封面图像；将所述质量评分最高的候选封面图像作为所述目标视频的最终封面图像。
与上面的方法实施例相对应，参见图5，本公开实施例还提供了一种视频封面生成装置50，包括：
解析模块501，用于对目标视频进行解析，得到多个解析图像。
目标视频是一段记载有音频和影像资料的视频文件，目标视频可以是任意格式的视频文件，例如，目标视频可以是mpg、mp4、rm、rmvb、wax格式的视频文件，也可以是其他格式的视频文件。
目标视频中包含有视频帧，视频帧是目标视频中所有视频图片的集合，例如，对于一个帧率为30fps的视频而言，按照正常的视频播放速度，1秒钟长度的视频中可以拆分出30个视频帧。当然基于实际的需要，也可以通过在30个视频帧中进行插帧的方式，获得更多的视频帧，或者，从30个视频帧中选择部分视频帧。
在对目标视频进行解析的过程中，为了提高解析的效率，可以对目标视频中的视频帧进行筛选，将符合筛选条件的视频帧作为解析图像。作为一个例子，可以对所述目标视频中包含的所有视频帧执行目标检测，通过目标检测，可以判断视频帧是否包含目标对象(例如，人)，当视频帧中包含目标对象时，则将包含所述目标对象的视频帧设置为所述解析图像。通过这种方式，能够进一步的提高多个解析图像的针对性。对视频帧进行目标检测可以采用多种针对图像的目标检测方法，在此对目标检测的方式不作限定。
聚类模块502，用于对所述多个解析图像进行聚类处理，得到多个候选封面图像。
解析图像中均含有目标对象，为了能够对包含目标对象的解析图像做进一步的筛选，可以对多个解析图像执行聚类处理，通过对图像进行聚类，可以在多个解析图像中选择一部分典型图像(多个候选封面图像)做进一步的处理。
具体的，可以预先对多个解析图像设置k类，从而进行聚类计算。首先， 在所述多个解析图像中选取k个样本点为初始聚类中心，记为z1(l),z2(l),……zk(l)，迭代序号l＝1；接下来，使用最近邻规则将所有样本分配到各聚类中心所代表的k类ωj(k)中，各类所包含的样本数为Nj(l)；接下来，通过计算各类的重心的方式，将计算得到的重心确定为新的聚类中心；最后，通过循环迭代的方式对于第j次迭代，判断zj(l+1)与zj(l)的值是否相同，当zj(l+1)≠zj(l)时继续迭代计算，当zj(l+1)＝zj(l)时，停止迭代计算。在迭代停止之后，可以通过最终聚类的结果，在每个聚类中选择一个解析图像，最终形成k个候选封面图像。
过滤模块503，用于基于对所述候选封面图像中包含的目标对象执行的关键点检测结果，对所述多个候选封面图像中包含的目标对象执行视线检测和睁眼检测，以过滤掉不符合预设要求的候选封面图像。
获取到多个候选封面图像之后，可以对候选封面图像中的目标对象执行关键点检测，通过关键点检测，可以得到目标图像头部的关键点，基于头部的关键点，可以对目标对象在候选封面图像中的视线和是否睁眼进行检测，从而将目标对象视线偏移或眼睛没有完全睁开的图像过滤掉。
具体的，可以获取对候选封面图像中包含的目标对象执行的关键点检测结果，基于这些关键点检测结果，确定出瞳孔关键点、眼球关键点及眼轮廓关键点；基于所述瞳孔关键点、眼球关键点及眼轮廓关键点，对所述目标对象的眼球执行建模操作，通过建模操作的结果，能够对所述目标对象执行视线检测和睁眼检测。
作为一种方式，作为一种方式，可以通过建模操作的结果使用CLNF模型检测出眼球的关键点，同时对眼睛进行三维建模，并在对眼睛完成三维建模之后，连接原点到瞳孔中心形成一条射线，计算其与眼球的交点，将眼球中心到交点方向的向量作为视线方向。从而进一步的判断目标对象的视线是否符合要求。
作为另外一种方式，可以通过建模操作的结果，计算外眼轮廓的宽高比，通过判断所述宽高比是否大于预设阈值来判断所述目标对象是否处于睁眼状态。
当通过视线检测和睁眼检测发现候选封面图像不符合要求时，可以将不符合要求的候选封面图像过滤掉，从而得到过滤后剩余的候选封面图像集合。
选择模块504，用于基于图像质量评价网络对过滤后剩余的候选封面图像的质量评分，从过滤后剩余的候选封面图像中选取目标图像作为所述目标视频的最终封面图像。
得到过滤后剩余的候选封面图像集合之后，可以从该集合中进一步的筛选一个图像，从而确定目标视频的最终封面图像。
作为一种方式，可以对所述多个候选封面图像中包含的目标对象执行视线检测和睁眼检测之前，通过使用预先训练的卷积神经网络对所有候选封面图像进行质量评分，质量评分的过程中，可以从图像画质、色彩、环境和表情等多个方面进行综合评价，从而对每个候选封面图像给出一个具体的质量评分。
在得到过滤后剩余的候选封面图像集合中每一个图像的质量评分之后，可以通过排序的方式，选择质量评分高的一个或多个候选封面图像作为该目标视频的最终封面图像。
图5所示装置可以对应的执行上述方法实施例中的内容，本实施例未详细描述的部分，参照上述方法实施例中记载的内容，在此不再赘述。
参见图6，本公开实施例还提供了一种电子设备60，该电子设备包括：
至少一个处理器；以及，
与该至少一个处理器通信连接的存储器；其中，
该存储器存储有可被该至少一个处理器执行的指令，该指令被该至少一个处理器执行，以使该至少一个处理器能够执行前述方法实施例中视频封面生成方法。
本公开实施例还提供了一种非暂态计算机可读存储介质，该非暂态计算机可读存储介质存储计算机指令，该计算机指令用于使该计算机执行前述方法实施例中。
本公开实施例还提供了一种计算机程序产品，该计算机程序产品包括存储在非暂态计算机可读存储介质上的计算程序，该计算机程序包括程序指令，当 该程序指令被计算机执行时，使该计算机执行前述方法实施例中的视频封面生成方法。
下面参考图6，其示出了适于用来实现本公开实施例的电子设备60的结构示意图。本公开实施例中的电子设备可以包括但不限于诸如移动电话、笔记本电脑、数字广播接收器、PDA(个人数字助理)、PAD(平板电脑)、PMP(便携式多媒体播放器)、车载终端(例如车载导航终端)等等的移动终端以及诸如数字TV、台式计算机等等的固定终端。图6示出的电子设备仅仅是一个示例，不应对本公开实施例的功能和使用范围带来任何限制。
如图6所示，电子设备60可以包括处理装置(例如中央处理器、图形处理器等)601，其可以根据存储在只读存储器(ROM)602中的程序或者从存储装置608加载到随机访问存储器(RAM)603中的程序而执行各种适当的动作和处理。在RAM 603中，还存储有电子设备60操作所需的各种程序和数据。处理装置601、ROM 602以及RAM 603通过总线604彼此相连。输入/输出(I/O)接口605也连接至总线604。
通常，以下装置可以连接至I/O接口605：包括例如触摸屏、触摸板、键盘、鼠标、图像传感器、麦克风、加速度计、陀螺仪等的输入装置606；包括例如液晶显示器(LCD)、扬声器、振动器等的输出装置607；包括例如磁带、硬盘等的存储装置608；以及通信装置609。通信装置609可以允许电子设备60与其他设备进行无线或有线通信以交换数据。虽然图中示出了具有各种装置的电子设备60，但是应理解的是，并不要求实施或具备所有示出的装置。可以替代地实施或具备更多或更少的装置。
特别地，根据本公开的实施例，上文参考流程图描述的过程可以被实现为计算机软件程序。例如，本公开的实施例包括一种计算机程序产品，其包括承载在计算机可读介质上的计算机程序，该计算机程序包含用于执行流程图所示的方法的程序代码。在这样的实施例中，该计算机程序可以通过通信装置609从网络上被下载和安装，或者从存储装置608被安装，或者从ROM 602被安装。在该计算机程序被处理装置601执行时，执行本公开实施例的方法中限定的上 述功能。
需要说明的是，本公开上述的计算机可读介质可以是计算机可读信号介质或者计算机可读存储介质或者是上述两者的任意组合。计算机可读存储介质例如可以是——但不限于——电、磁、光、电磁、红外线、或半导体的***、装置或器件，或者任意以上的组合。计算机可读存储介质的更具体的例子可以包括但不限于：具有一个或多个导线的电连接、便携式计算机磁盘、硬盘、随机访问存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、光纤、便携式紧凑磁盘只读存储器(CD-ROM)、光存储器件、磁存储器件、或者上述的任意合适的组合。在本公开中，计算机可读存储介质可以是任何包含或存储程序的有形介质，该程序可以被指令执行***、装置或者器件使用或者与其结合使用。而在本公开中，计算机可读信号介质可以包括在基带中或者作为载波一部分传播的数据信号，其中承载了计算机可读的程序代码。这种传播的数据信号可以采用多种形式，包括但不限于电磁信号、光信号或上述的任意合适的组合。计算机可读信号介质还可以是计算机可读存储介质以外的任何计算机可读介质，该计算机可读信号介质可以发送、传播或者传输用于由指令执行***、装置或者器件使用或者与其结合使用的程序。计算机可读介质上包含的程序代码可以用任何适当的介质传输，包括但不限于：电线、光缆、RF(射频)等等，或者上述的任意合适的组合。
上述计算机可读介质可以是上述电子设备中所包含的；也可以是单独存在，而未装配入该电子设备中。
上述计算机可读介质承载有一个或者多个程序，当上述一个或者多个程序被该电子设备执行时，使得该电子设备：获取至少两个网际协议地址；向节点评价设备发送包括所述至少两个网际协议地址的节点评价请求，其中，所述节点评价设备从所述至少两个网际协议地址中，选取网际协议地址并返回；接收所述节点评价设备返回的网际协议地址；其中，所获取的网际协议地址指示内容分发网络中的边缘节点。
或者，上述计算机可读介质承载有一个或者多个程序，当上述一个或者多 个程序被该电子设备执行时，使得该电子设备：接收包括至少两个网际协议地址的节点评价请求；从所述至少两个网际协议地址中，选取网际协议地址；返回选取出的网际协议地址；其中，接收到的网际协议地址指示内容分发网络中的边缘节点。
可以以一种或多种程序设计语言或其组合来编写用于执行本公开的操作的计算机程序代码，上述程序设计语言包括面向对象的程序设计语言—诸如Java、Smalltalk、C++，还包括常规的过程式程序设计语言—诸如“C”语言或类似的程序设计语言。程序代码可以完全地在用户计算机上执行、部分地在用户计算机上执行、作为一个独立的软件包执行、部分在用户计算机上部分在远程计算机上执行、或者完全在远程计算机或服务器上执行。在涉及远程计算机的情形中，远程计算机可以通过任意种类的网络——包括局域网(LAN)或广域网(WAN)—连接到用户计算机，或者，可以连接到外部计算机(例如利用因特网服务提供商来通过因特网连接)。
附图中的流程图和框图，图示了按照本公开各种实施例的***、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段、或代码的一部分，该模块、程序段、或代码的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。也应当注意，在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个接连地表示的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合，可以用执行规定的功能或操作的专用的基于硬件的***来实现，或者可以用专用硬件与计算机指令的组合来实现。
描述于本公开实施例中所涉及到的单元可以通过软件的方式实现，也可以通过硬件的方式来实现。其中，单元的名称在某种情况下并不构成对该单元本身的限定，例如，第一获取单元还可以被描述为“获取至少两个网际协议地址的单元”。
应当理解，本公开的各部分可以用硬件、软件、固件或它们的组合来实现。
以上所述，仅为本公开的具体实施方式，但本公开的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本公开揭露的技术范围内，可轻易想到的变化或替换，都应涵盖在本公开的保护范围之内。因此，本公开的保护范围应以权利要求的保护范围为准。

Claims (12)

  1. 一种视频封面生成方法，其特征在于，包括：
    对目标视频进行解析，得到多个解析图像；
    对所述多个解析图像进行聚类处理，得到多个候选封面图像；
    基于对所述候选封面图像中包含的目标对象执行的关键点检测结果，对所述多个候选封面图像中包含的目标对象执行视线检测和睁眼检测，以过滤掉不符合预设要求的候选封面图像；
    基于图像质量评价网络对过滤后剩余的候选封面图像的质量评分，从过滤后剩余的候选封面图像中选取目标图像作为所述目标视频的最终封面图像。
  2. 根据权利要求1所述的方法，其特征在于，所述对目标视频进行解析，得到多个解析图像，包括：
    对所述目标视频中包含的所有视频帧执行目标检测；
    基于目标检测的结果，判断组成目标视频的视频帧中是否包含所述目标对象；
    若是，则将包含所述目标对象的视频帧设置为所述解析图像。
  3. 根据权利要求1所述的方法，其特征在于，所述对所述多个解析图像进行聚类处理，得到多个候选封面图像，包括：
    对所述多个解析图像执行k类的聚类计算；
    在每个聚类中选择一张符合预设条件的图像作为所述候选封面图像。
  4. 根据权利要求3所述的方法，其特征在于，所述对所述多个解析图像执行k类的聚类计算，包括：
    在所述多个解析图像中选取k个样本点为初始聚类中心，记为z1(l),z2(l),……zk(l)，迭代序号l＝1；
    使用最近邻规则将所有样本分配到各聚类中心所代表的k类ωj(k)中，各类所包含的样本数为Nj(l)；
    计算各类的重心，将计算得到的重心确定为新的聚类中心；
    对于第j次迭代，判断zj(l+1)与zj(l)的值是否相同，当zj(l+1)≠zj(l)时继续 迭代计算，当zj(l+1)＝zj(l)时，停止迭代计算。
  5. 根据权利要求1所述的方法，其特征在于，所述对所述多个候选封面图像中包含的目标对象执行视线检测和睁眼检测之前，所述方法还包括：
    利用预先设置的卷积神经网络对所述多个候选封面图像执行质量评价；
    基于所述质量评价的结果，确定所述多个候选封面图像中每一个图像的质量评分。
  6. 根据权利要求1所述的方法，其特征在于，所述对所述多个候选封面图像中包含的目标对象执行视线检测和睁眼检测，包括：
    获取对候选封面图像中包含的目标对象执行的关键点检测结果；
    基于所述关键点检测结果，确定出瞳孔关键点、眼球关键点及眼轮廓关键点；
    基于所述瞳孔关键点、眼球关键点及眼轮廓关键点，对所述目标对象的眼球执行建模操作；
    基于建模操作的结果，来对所述目标对象执行视线检测和睁眼检测。
  7. 根据权利要求6所述的方法，其特征在于，所述基于建模操作的结果，来对所述目标对象执行视线检测和睁眼检测，包括：
    使用CLNF模型检测出眼球的关键点；
    对眼睛进行三维建模；
    连接原点到瞳孔中心形成一条射线，计算其与眼球的交点，将眼球中心到交点方向的向量作为视线方向。
  8. 根据权利要求6所述的方法，其特征在于，所述基于建模操作的结果，来对所述目标对象执行视线检测和睁眼检测，包括：
    通过建模操作的结果，计算外眼轮廓的宽高比；
    通过判断所述宽高比是否大于预设阈值来判断所述目标对象是否处于睁眼状态。
  9. 根据权利要求1所述的方法，其特征在于，所述基于图像质量评价网络对过滤后剩余的候选封面图像的质量评分，从过滤后剩余的候选封面图像中选 取目标图像作为所述目标视频的最终封面图像，包括：
    从过滤后剩余的候选封面图像中选取质量评分最高的候选封面图像；
    将所述质量评分最高的候选封面图像作为所述目标视频的最终封面图像。
  10. 一种视频封面生成装置，其特征在于，包括：
    解析模块，用于对目标视频进行解析，得到多个解析图像；
    聚类模块，用于对所述多个解析图像进行聚类处理，得到多个候选封面图像；
    过滤模块，用于基于对所述候选封面图像中包含的目标对象执行的关键点检测结果，对所述多个候选封面图像中包含的目标对象执行视线检测和睁眼检测，以过滤掉不符合预设要求的候选封面图像；
    选择模块，用于基于图像质量评价网络对过滤后剩余的候选封面图像的质量评分，从过滤后剩余的候选封面图像中选取目标图像作为所述目标视频的最终封面图像。
  11. 一种电子设备，其特征在于，所述电子设备包括：
    至少一个处理器；以及，
    与所述至少一个处理器通信连接的存储器；其中，
    所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行前述任一权利要求1-9所述的视频封面生成方法。
  12. 一种非暂态计算机可读存储介质，该非暂态计算机可读存储介质存储计算机指令，该计算机指令用于使该计算机执行前述任一权利要求1-9所述的视频封面生成方法。
PCT/CN2020/096719 2019-07-11 2020-06-18 视频封面生成方法、装置及电子设备 WO2021004247A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910622565.6A CN110381368A (zh) 2019-07-11 2019-07-11 视频封面生成方法、装置及电子设备
CN201910622565.6 2019-07-11

Publications (1)

Publication Number Publication Date
WO2021004247A1 true WO2021004247A1 (zh) 2021-01-14

Family

ID=68252598

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/096719 WO2021004247A1 (zh) 2019-07-11 2020-06-18 视频封面生成方法、装置及电子设备

Country Status (2)

Country Link
CN (1) CN110381368A (zh)
WO (1) WO2021004247A1 (zh)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113723242A (zh) * 2021-08-20 2021-11-30 湖南全航信息通信有限公司 基于视频终端的视觉测谎方法、电子设备及介质
CN114449362A (zh) * 2022-03-17 2022-05-06 腾讯科技（上海）有限公司 视频封面的选取方法、装置、设备及存储介质
CN115134677A (zh) * 2022-05-30 2022-09-30 一点灵犀信息技术(广州)有限公司 视频封面选择方法、装置、电子设备以及计算机存储介质
WO2022206729A1 (zh) * 2021-04-01 2022-10-06 影石创新科技股份有限公司 视频封面选择方法、装置、计算机设备和存储介质
WO2023124874A1 (zh) * 2021-12-28 2023-07-06 北京字节跳动网络技术有限公司 问答结果的搜索方法、装置、设备及存储介质
CN117689782A (zh) * 2024-02-02 2024-03-12 腾讯科技（深圳）有限公司 一种生成海报图像的方法、装置、设备及存储介质

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110381368A (zh) * 2019-07-11 2019-10-25 北京字节跳动网络技术有限公司 视频封面生成方法、装置及电子设备
CN110929070A (zh) * 2019-12-09 2020-03-27 北京字节跳动网络技术有限公司 图像处理方法、装置、电子设备及存储介质
CN111062314B (zh) * 2019-12-13 2021-11-02 腾讯科技（深圳）有限公司 图像选取方法、装置、计算机可读存储介质及电子设备
CN111464833B (zh) * 2020-03-23 2023-08-04 腾讯科技（深圳）有限公司 目标图像生成方法、目标图像生成装置、介质及电子设备
CN111491182B (zh) * 2020-04-23 2022-03-29 百度在线网络技术（北京）有限公司 用于视频封面存储解析的方法和装置
CN111859022A (zh) * 2020-07-07 2020-10-30 咪咕文化科技有限公司 封面生成方法、电子设备及计算机可读存储介质
CN112749711B (zh) * 2020-08-04 2023-08-25 腾讯科技（深圳）有限公司 视频获取方法和装置及存储介质
CN112308864A (zh) * 2020-10-27 2021-02-02 北京达佳互联信息技术有限公司 一种图像处理方法、装置、电子设备及存储介质

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100121852A1 (en) * 2008-11-11 2010-05-13 Samsung Electronics Co., Ltd Apparatus and method of albuming content
CN107392159A (zh) * 2017-07-27 2017-11-24 竹间智能科技（上海）有限公司 一种面部专注度检测***及方法
CN107832725A (zh) * 2017-11-17 2018-03-23 北京奇虎科技有限公司 基于评价指标的视频封面提取方法及装置
CN108875524A (zh) * 2018-01-02 2018-11-23 北京旷视科技有限公司 视线估计方法、装置、***和存储介质
CN109002812A (zh) * 2018-08-08 2018-12-14 北京未来媒体科技股份有限公司 一种智能识别视频封面的方法及装置
CN109165301A (zh) * 2018-09-13 2019-01-08 北京字节跳动网络技术有限公司 视频封面选择方法、装置和计算机可读存储介质
CN109635747A (zh) * 2018-12-14 2019-04-16 央视国际网络无锡有限公司 视频封面自动抽取方法及装置
CN110381368A (zh) * 2019-07-11 2019-10-25 北京字节跳动网络技术有限公司 视频封面生成方法、装置及电子设备

Family Cites Families (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110047163A1 (en) * 2009-08-24 2011-02-24 Google Inc. Relevance-Based Image Selection
CN102830793B (zh) * 2011-06-16 2017-04-05 北京三星通信技术研究有限公司 视线跟踪方法和设备
KR20130134546A (ko) * 2012-05-31 2013-12-10 삼성전자주식회사 동영상의 섬네일 이미지 생성 방법 및 그 전자 장치
CN105894031A (zh) * 2016-03-31 2016-08-24 青岛海信移动通信技术股份有限公司 一种照片选择的方法及装置
CN106529406B (zh) * 2016-09-30 2020-02-07 广州华多网络科技有限公司 一种获取视频摘要图像的方法和装置
CN108733676A (zh) * 2017-04-14 2018-11-02 合信息技术（北京）有限公司 视频缩略图的提取方法及装置
CN107147939A (zh) * 2017-05-05 2017-09-08 百度在线网络技术（北京）有限公司 用于调整视频直播封面的方法和装置
CN108108684B (zh) * 2017-12-15 2020-07-17 杭州电子科技大学 一种融合视线检测的注意力检测方法
CN108833939A (zh) * 2018-06-20 2018-11-16 北京优酷科技有限公司 生成视频的海报的方法及装置
CN108833942A (zh) * 2018-06-28 2018-11-16 北京达佳互联信息技术有限公司 视频封面选取方法、装置、计算机设备及存储介质
CN109583292A (zh) * 2018-10-11 2019-04-05 杭州电子科技大学 一种视线区域检测方法
CN110555426A (zh) * 2019-09-11 2019-12-10 北京儒博科技有限公司 视线检测方法、装置、设备及存储介质

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100121852A1 (en) * 2008-11-11 2010-05-13 Samsung Electronics Co., Ltd Apparatus and method of albuming content
CN107392159A (zh) * 2017-07-27 2017-11-24 竹间智能科技（上海）有限公司 一种面部专注度检测***及方法
CN107832725A (zh) * 2017-11-17 2018-03-23 北京奇虎科技有限公司 基于评价指标的视频封面提取方法及装置
CN108875524A (zh) * 2018-01-02 2018-11-23 北京旷视科技有限公司 视线估计方法、装置、***和存储介质
CN109002812A (zh) * 2018-08-08 2018-12-14 北京未来媒体科技股份有限公司 一种智能识别视频封面的方法及装置
CN109165301A (zh) * 2018-09-13 2019-01-08 北京字节跳动网络技术有限公司 视频封面选择方法、装置和计算机可读存储介质
CN109635747A (zh) * 2018-12-14 2019-04-16 央视国际网络无锡有限公司 视频封面自动抽取方法及装置
CN110381368A (zh) * 2019-07-11 2019-10-25 北京字节跳动网络技术有限公司 视频封面生成方法、装置及电子设备

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2022206729A1 (zh) * 2021-04-01 2022-10-06 影石创新科技股份有限公司 视频封面选择方法、装置、计算机设备和存储介质
CN113723242A (zh) * 2021-08-20 2021-11-30 湖南全航信息通信有限公司 基于视频终端的视觉测谎方法、电子设备及介质
CN113723242B (zh) * 2021-08-20 2024-04-26 湖南全航信息通信有限公司 基于视频终端的视觉测谎方法、电子设备及介质
WO2023124874A1 (zh) * 2021-12-28 2023-07-06 北京字节跳动网络技术有限公司 问答结果的搜索方法、装置、设备及存储介质
CN114449362A (zh) * 2022-03-17 2022-05-06 腾讯科技（上海）有限公司 视频封面的选取方法、装置、设备及存储介质
CN114449362B (zh) * 2022-03-17 2023-08-22 腾讯科技（上海）有限公司 视频封面的选取方法、装置、设备及存储介质
CN115134677A (zh) * 2022-05-30 2022-09-30 一点灵犀信息技术(广州)有限公司 视频封面选择方法、装置、电子设备以及计算机存储介质
CN117689782A (zh) * 2024-02-02 2024-03-12 腾讯科技（深圳）有限公司 一种生成海报图像的方法、装置、设备及存储介质
CN117689782B (zh) * 2024-02-02 2024-05-28 腾讯科技（深圳）有限公司 一种生成海报图像的方法、装置、设备及存储介质

Also Published As

Publication number Publication date
CN110381368A (zh) 2019-10-25

Similar Documents

Publication Publication Date Title
WO2021004247A1 (zh) 视频封面生成方法、装置及电子设备
WO2021008223A1 (zh) 信息的确定方法、装置及电子设备
CN108012173B (zh) 一种内容识别方法、装置、设备和计算机存储介质
CN110399848A (zh) 视频封面生成方法、装置及电子设备
JP2023547917A (ja) 画像分割方法、装置、機器および記憶媒体
WO2020248900A1 (zh) 全景视频的处理方法、装置及存储介质
WO2020228383A1 (zh) 口型生成方法、装置及电子设备
WO2021135319A1 (zh) 基于深度学习的文案生成方法、装置及电子设备
CN110047121B (zh) 端到端的动画生成方法、装置及电子设备
WO2020220809A1 (zh) 目标对象的动作识别方法、装置和电子设备
WO2020211573A1 (zh) 用于处理图像的方法和装置
WO2022037419A1 (zh) 音频内容识别方法、装置、设备和计算机可读介质
WO2020186934A1 (zh) 包含动态背景的动画生成方法、装置及电子设备
CN113723087B (zh) 信息处理方法、装置、设备、可读存储介质及产品
CN110035271B (zh) 保真图像生成方法、装置及电子设备
CN110930220A (zh) 一种显示方法、装置、终端设备及介质
WO2023011318A1 (zh) 媒体文件处理方法、装置、设备、可读存储介质及产品
CN111738316B (zh) 零样本学习的图像分类方法、装置及电子设备
CN111222509A (zh) 目标检测方法、装置及电子设备
CN110287816B (zh) 车门动作检测方法、装置和计算机可读存储介质
CN110060324B (zh) 图像渲染方法、装置及电子设备
CN110335237B (zh) 用于生成模型的方法、装置和用于识别图像的方法、装置
CN109857244B (zh) 一种手势识别方法、装置、终端设备、存储介质及vr眼镜
CN111626990A (zh) 目标检测框处理方法、装置及电子设备
CN110969159A (zh) 图像识别方法、装置及电子设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20837723

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 22/04/2022)

122 Ep: pct application non-entry in european phase

Ref document number: 20837723

Country of ref document: EP

Kind code of ref document: A1