KR102742878B1 - Apparatus and method for editing an image data using an artificial intelligence automatically in the image editing apparatus - Google Patents

Abstract

According to one embodiment of the present invention, a device and method for automatically editing image data using artificial intelligence in an image editing device include an image collection unit for collecting and storing a plurality of learning image data; an image input unit for receiving and storing a plurality of original image data; a scene learning database for storing a plurality of scene learning image data by image type; an editing learning database for storing a plurality of edited learning image data by image type; an artificial intelligence engine learning unit for selecting a plurality of learning image data corresponding to a specific image type input by a user from among the learning image data, and for training an artificial intelligence engine using the selected learning image data and a pre-designated scene learning parameter and a pre-designated editing learning parameter; an image synchronization unit for synchronizing a reference point of the original image data based on a pre-designated specific sound; a scene analysis unit for analyzing a plurality of scenes from the synchronized original image data using the scene learning image data; and an editing analysis unit for applying the edited learning image data to an artificial intelligence program using the artificial intelligence engine to select a plurality of scenes from the analyzed scenes, and editing the selected scenes to generate edited image data.

Description

{Apparatus and method for editing an image data using an artificial intelligence automatically in the image editing apparatus}

The present invention relates to a video editing device, and more particularly, to a device and method for automatically editing video data using artificial intelligence in a video editing device.

With the development of the Internet and networks, media production is now actively taking place not only through mass media broadcasting through large broadcasting stations, but also through individual media broadcasting. This expansion of media broadcasting has led to a rapid increase in video editing.

In particular, in order to spread a video through media such as broadcasting or SNS (social network service), the video must be edited according to the purpose of the video. Here, the purpose of the video includes advertising or a summary of the highlights of the broadcast.

However, since this video editing is done manually by users (e.g., editors), there was a problem that it took a lot of time and effort. For example, if the video editing is cross-editing, even an experienced editor can take 5 to 10 hours or more to produce a video of less than 5 minutes through cross-editing. Here, a cross-edited video is a secondary work in which an editor arbitrarily edits several stage videos of an idol group to create a single video in which the flow of the song and choreography is continuous, and it is very popular on video providing platforms such as YouTube.

And there was a problem that the quality of the results produced through video editing had a large difference depending on the skill of the editor.

Therefore, the need for a solution to solve these problems has arisen.

One embodiment of the present invention proposes a device and method for automatically editing image data in a short time and with little effort in an image editing device.

And one embodiment of the present invention proposes a device and method for automatically editing image data so that the result has a constant quality in an image editing device.

According to one embodiment of the present invention, a device for automatically editing image data using artificial intelligence in an image editing device includes an image collection unit for collecting and storing a plurality of learning image data; an image input unit for receiving and storing a plurality of original image data; a scene learning database for storing a plurality of scene learning image data by image type; an editing learning database for storing a plurality of edited learning image data by image type; an artificial intelligence engine learning unit for selecting a plurality of learning image data corresponding to a specific image type input by a user from among the learning image data, and for training an artificial intelligence engine using the selected learning image data and a pre-designated scene learning parameter and a pre-designated editing learning parameter; an image synchronization unit for synchronizing a reference point of the original image data based on a pre-designated specific sound; a scene analysis unit for analyzing a plurality of scenes from the synchronized original image data using the scene learning image data; and an editing analysis unit for applying the edited learning image data to an artificial intelligence program using the artificial intelligence engine to select a plurality of scenes from the analyzed scenes, and editing the selected scenes to generate edited image data.

According to one embodiment of the present invention, a method for automatically editing image data using artificial intelligence in an image editing device includes: an image collection unit collecting and storing a plurality of training image data; an artificial intelligence engine learning unit selecting a plurality of training image data corresponding to a specific image type input by a user from among the training image data; an artificial intelligence engine learning unit training an artificial intelligence engine using the selected training image data and a pre-designated scene learning parameter and a pre-designated editing learning parameter; an image input unit receiving and storing a plurality of original image data; an image synchronization unit synchronizing a reference point of the original image data based on a pre-designated specific sound; a scene analysis unit analyzing a plurality of scenes from the synchronized original image data using the scene learning image data; an editing analysis unit applying the editing learning image data to an artificial intelligence program using the artificial intelligence engine to select a plurality of scenes from among the analyzed scenes; and an editing analysis unit editing the selected scenes to generate edited image data.

One embodiment of the present invention automatically edits image data using artificial intelligence in an image editing device, thereby enabling the editing of image data in a short time and with little effort.

And one embodiment of the present invention can automatically edit image data using artificial intelligence in an image editing device so that the result has a consistent quality.

In addition, the effects that can be obtained or expected by the embodiments of the present invention will be directly or implicitly disclosed in the detailed description of the embodiments of the present invention. That is, the various effects expected according to the embodiments of the present invention will be disclosed in the detailed description to be described later.

FIG. 1 is a block diagram of an image editing device according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating an operation of synchronizing reference points of a plurality of image data in an image synchronization unit according to one embodiment of the present invention.
FIG. 3 is a diagram illustrating an operation of editing a plurality of image data in an editing analysis unit according to an embodiment of the present invention.
FIG. 4 is a flowchart of automatically editing image data in an image editing device according to an embodiment of the present invention.
FIG. 5 is a flowchart of learning an artificial intelligence engine in an artificial intelligence engine learning unit according to one embodiment of the present invention.

The terms used in this specification will be briefly explained, and the present invention will be described in detail.

The terms used in the embodiments of the present invention are selected from the most widely used general terms possible while considering the functions of the present invention, but this may vary depending on the intention of engineers working in the field, precedents, the emergence of new technologies, etc. In addition, in certain cases, there are terms arbitrarily selected by the applicant, and in this case, the meanings thereof will be described in detail in the description of the corresponding invention. Therefore, the terms used in the present invention should be defined based on the meanings of the terms and the overall contents of the present invention, rather than simply the names of the terms.

The embodiments of the present invention can be variously transformed and have various embodiments, and specific embodiments are illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the scope to specific embodiments, and it should be understood that it includes all transformations, equivalents, or substitutes included in the invention's idea and technical scope. In describing the embodiments, if it is judged that a specific description of a related known technology may obscure the main point, the detailed description is omitted.

The terms first, second, etc. may be used to describe various components, but the components should not be limited by the terms. The terms are used only to distinguish one component from another.

Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, the terms "comprise" or "consist of" and the like are intended to specify the presence of a feature, number, step, operation, component, part or combination thereof described in the specification, but should be understood to not preclude the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts or combinations thereof.

In an embodiment of the present invention, a 'module' or 'part' performs at least one function or operation, and may be implemented by hardware or software, or by a combination of hardware and software. In addition, a plurality of 'modules' or a plurality of 'parts' may be integrated into at least one module and implemented by at least one processor (not shown), except for a 'module' or 'part' that needs to be implemented by specific hardware.

In the embodiments of the present invention, when a part is said to be "connected" to another part, this includes not only the case where it is "directly connected" but also the case where it is "electrically connected" with another element in between. In addition, when a part is said to "include" a certain component, this does not mean that the other component is excluded, but that the other component can be further included, unless specifically stated otherwise.

Hereinafter, with reference to the attached drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily practice them. However, the present invention may be implemented in various different forms and is not limited to the embodiments described herein. In addition, in order to clearly describe the present invention in the drawings, parts that are not related to the description are omitted, and similar parts are assigned similar drawing reference numerals throughout the specification.

FIG. 1 is a block diagram of an image editing device according to an embodiment of the present invention.

Referring to FIG. 1, the image editing device (101) includes an image collection unit (103), an artificial intelligence (AI) engine learning unit (105), a scene learning database (DataBase, hereinafter referred to as 'DB') (107), an editing learning database (109), an image input unit (111), an image synchronization unit (113), a scene analysis unit (115), and an editing analysis unit (117).

Looking at each component, the video collection unit (103) collects and stores a plurality of externally produced learning video data (125). For example, the learning video data are already edited video data, and can be classified and stored according to the type of the learning video data. For example, the type of the learning video data can be classified according to the genre of the video data. For example, the genre of the video data can include entertainment video data, performance video data (for example, stage video data, etc.), review video data (for example, car, real estate, or toy review video data, etc.), sports video data (for example, on-site or broadcast sports video data, etc.), highlight video data (for example, video, drama, sports (for example, scoring scene highlight, etc.), entertainment or documentary highlight video data, etc.). For example, the selected learning video data can be learning video data having the same length.

The artificial intelligence engine learning unit (103) learns the artificial intelligence engine by using classified image data selected by the user from among a plurality of learning image data (125). For example, the artificial intelligence engine may be software that acts as a brain in an artificial intelligence program. For example, the artificial intelligence program may be software that provides a function of editing a plurality of image data to create a single image data. For example, the artificial intelligence program may be a ResNext-101 model.

To explain in more detail, the artificial intelligence engine learning unit (103) receives an image type for learning of the artificial intelligence engine from the user. Then, the artificial intelligence engine learning unit (103) selects a plurality of learning image data corresponding to the input image type from among a plurality of learning image data.

And the artificial intelligence engine learning unit (103) trains the artificial intelligence engine by applying pre-designated scene learning parameters to the selected learning image data. For example, the scene learning parameters may be for extracting characteristics of good scenes from the selected learning image data.

For example, scene learning parameters may include at least one of number of speakers, object movement, scene composition, camera movement, and audio syllables.

For example, the number of speakers may include selecting a scene with one speaker when there is one speaker, and selecting a scene with multiple speakers when there are two or more speakers. For example, object movement may include learning a rule that divides a section where the speed at which an object moves on the screen changes into a unit scene, learning a scene where an object moves a certain distance away from the center of the screen, and learning a scene where an object with different movement exists among multiple moving objects.

For example, the screen composition may include a screen composition according to a central object and a screen composition according to a camera angle. For example, the screen composition according to a central object may be divided into a person-centered composition and a non-person-centered composition. For example, in the case of a person-centered composition, the screen composition may include a bust shot, a full shot, or a close-up (e.g., centered on a face). For example, in the case of a non-person-centered composition, the screen composition may vary depending on whether an object (e.g., a non-person) is included, and if an object is included, it may include giving priority to a full shot with the object at the center of the screen. For example, the screen composition according to a camera angle may select a scene depending on whether it is from a special angle (e.g., a bird's eye view or a bird's eye view).

For example, the camera movement may include whether to pan, whether to tilt, and whether to zoom. For example, whether to pan may be to determine the judgment criteria for the start and end points of the pan by considering a pre-specified appropriate camera panning time when the camera angle changes more than 5 degrees to the left or right. For example, whether to tilt may be to determine the judgment criteria for the start and end points of the tilt by considering a pre-specified appropriate camera tilting time when the camera angle changes more than 2 degrees to the up or down. For example, whether to zoom may be to learn the degree of change in the size of the zoom target object increasing or learning the degree of change in the size of the zoom target object decreasing.

For example, audio syllables may include audio loudness changes, syllables being grouped, or a speaker speaking a particular sentence. For example, audio loudness changes may include audio loudness changes or silences exceeding a predetermined amount of time. For example, a particular sentence may include at least one of "Shall we take a look?" and "Let's take a look."

And the artificial intelligence engine learning unit (103) generates at least one image data (hereinafter referred to as 'scene learning image data') corresponding to at least one scene that satisfies the scene learning parameters among the selected learning image data. And the artificial intelligence engine learning unit (103) stores the generated scene learning image data in the scene learning database (107).

And the artificial intelligence engine learning unit (103) trains the artificial intelligence engine by applying pre-specified editing learning parameters to selected learning image data. For example, the editing learning parameters may be for selecting multiple scenes that satisfy the editing learning parameters among a plurality of scenes and editing the selected scenes.

For example, the editing learning parameters may include at least one of scene transition, audio transition, dialogue recognition, action recognition, and background music position.

For example, scene transitions may include screen complexity between previous and next scenes, degree of motion within the screen, changes in lighting within the screen, changes in screen color scheme, changes in time of day within the screen (e.g., night or day), movement and size of objects (people or objects) within the screen, and screen transition techniques (e.g., fades, wipes, dissolves). For example, speech recognition may include pre-learning matching between people and audio, and recognizing speakers among objects on the screen.

For example, action recognition can include detecting bouncing motions within a screen and recognizing facial expressions. Detecting bouncing motions within a screen can be distinguishing between motion changes that can be distinguished in a sequence of scenes. Recognizing facial expressions can be recognizing and distinguishing between the speaker's facial expressions, such as laughter, crying, sadness, and joy.

For example, the background music location may include distinguishing at least one of: distinguishing a section where background music is inserted, distinguishing a change in the movement of an object in the section where background music is inserted, and distinguishing a difference in the dynamics and flow of the screen in the section where background music is inserted.

And the artificial intelligence engine learning unit (103) generates at least one image data (hereinafter referred to as “editing learning image data”) corresponding to at least one editing point that satisfies the editing learning parameter among the selected learning image data. And the artificial intelligence engine learning unit (103) stores the generated editing learning image data in the editing learning database (109).

The scene learning database (107) stores at least one scene learning image data. For example, the scene learning database (109) can store a plurality of scene learning image data by type of image data.

The editing learning database (109) stores at least one editing learning image data. For example, the editing learning database (109) can store a plurality of editing learning image data by type of image data.

The image input unit (111) receives and stores a plurality of original image data (123) from the user. For example, the original image data may be a plurality of image data captured by a plurality of image capturing cameras. At this time, the image input unit (111) may receive a plurality of types of original image data from the user.

The image synchronization unit (113) synchronizes the reference time of a plurality of original image data (123). For example, the image synchronization unit (113), as illustrated in FIG. 2, may perform sound synchronization by designating a pre-designated sound included in the original image data (123) as a reference time, and setting specific frames (e.g., xxxxx year yy month zz day aa hour bb minute cc second dd frame) of the designated reference time as starting longitudinal time codes (hereinafter referred to as 'LTC'). For example, the synchronization sound may be a clapping sound. For example, one scene of the image data may consist of 30 frames. At this time, the image synchronization unit (113) may separately use a master recording track or may use an audio track of a specific image capturing camera among a plurality of image capturing cameras as a master audio track. And the video synchronization unit (113) can perform video synchronization by designating the frames of the original video data (123) as the same point in time as the start vertical time codes of the original video data (e.g., dd frame, dd day, yy month, zz day, aa hour, bb minute, cc second, xxxx year). And the video synchronization unit (113) can remove a silent section before the start of an audio track and provide a frame margin after the end of an audio track.

The scene analysis unit (115) analyzes a plurality of scenes from a plurality of synchronized original image data using a pre-trained model. For example, the pre-trained model can represent a pre-trained model using data that is similar to a problem that the user wants to solve and is large in size. For example, the pre-trained model can be scene learning image data.

To explain in more detail, the scene analysis unit (115) receives a plurality of synchronized original image data from the image synchronization unit (113). Then, the scene analysis unit (115) passes the synchronized original image data through a pre-learned model and extracts a plurality of features corresponding to a plurality of scenes from the synchronized original image data. For example, the pre-learned model may be a plurality of scene learning image data having the same type as the plurality of original image data.

And the scene analysis unit (115) performs a concat process on each of the extracted features corresponding to multiple scenes to convert the extracted features into multiple data values. For example, the concat process may be a process of combining strings for the extracted features to generate specific data values. And the scene analysis unit (115) outputs the converted data values corresponding to multiple scenes to the editing analysis unit (117).

The editing analysis unit (117) selects multiple scenes from among a plurality of scenes by applying the data values converted to the artificial intelligence program. Then, the editing analysis unit (117) automatically edits the selected scenes based on a plurality of editing learning image data to generate one editing image data (119) and editing information (121). Thereafter, the editing analysis unit (117) stores the editing image data (119) and the editing information (121) in a database (not shown) or outputs them through a display unit (not shown). For example, the editing information (121) may include time information of the start and end points of the scene, editing order information, and editing recommendation accuracy information.

To explain in more detail, the editing analysis unit (117) inputs the converted data values into the artificial intelligence engine model. For example, the artificial intelligence engine model may be a learning engine configured by applying an artificial intelligence algorithm developed by the applicant. Then, the editing analysis unit (117) measures each parameter of the correct video class by type (for example, the number of speakers, the size of the object movement, the size of the camera angle change, the change in audio level, etc.) and generates and stores the result as editing recommendation accuracy information. Then, the editing analysis unit (117) calculates the measured actual data and the cross-entropy loss value of the parameter values for each input video scene. For example, the ground truth values represent the characteristic values of each scene for each parameter, and the cross entropy loss can represent the accuracy of editing recommendations.

For example, learning content can include whether or not to adopt a scene, the length of the scene composition, and editing points (e.g., scene transitions). For example, if the editing is a cross-editing of a music show featuring a specific idol group, the editing points can be divided into sections of a song sung by a specific idol group.

For example, the learning content can consist of gesture recognition. For example, gesture recognition can include speech of a person, recognition of movement (e.g., facial expression, gesture, movement, dance, etc.), recognition of the beginning and/or end of a unit scene, matching between voice and person (achieved through short-term learning).

For example, the editing analysis unit (117) can perform editing using pre-specified editing rules. For example, the pre-specified editing rules can vary depending on the number of video capturing cameras.

For example, when producing a video of a regular broadcast program featuring a specific idol group, since all cameras and cameras for each member of the idol group are assigned to the regular broadcast, the number of video shooting cameras may be four or more. In this case, the editing rules may include giving priority to the entire shot shot by all cameras (applying to the beginning and/or the end), using the entire shot when there are two or more voices in one scene, and recognizing the person's voice in one scene and pointing out and using the dedicated shot shot by the person's camera. For example, the editing analysis unit (117) may edit synchronized original video data using these editing rules, thereby configuring the beginning and the end of the edited video data as the entire shot, and configuring the remaining middle part as the member-specific shot and the right shot.

As another example, if a video is produced by an individual or a small group that is mainly active on SNS, the number of video cameras may be three or less. In this case, the editing rules may include excluding silence cuts (+/- frame margin), organizing the editing sequence based on LTC (absolute value) (screens can be taken from multiple original video data and used), automatically recognizing subtitles (according to the time interval), designating master audio (for example, the audio of the camera that shoots the Main Creator Shot can be designated as the master audio), and automatically inserting a title clip and an ending clip into the edited video data. For example, the editing analysis unit (117) may edit synchronized original video data using these editing rules to configure the beginning and end parts of the edited video data as the Main Creator Shot, and the remaining middle part as the Object Zoom Shot and the Right Shot, as in the graph 303 of FIG. 3. At this time, the editing analysis unit (117) provides an XML (eXtensible Markup Language) output function for edited video data in preparation for additional editing using a non-linear editing system (hereinafter referred to as 'NLE').

Through this configuration, one embodiment of the present invention can automatically edit image data using artificial intelligence in an image editing device, thereby editing image data in a short time and with little effort. And one embodiment of the present invention can automatically edit image data using artificial intelligence in an image editing device, thereby editing image data so that the result has a consistent quality.

FIG. 4 is a flowchart of automatically editing image data in an image editing device (101) according to one embodiment of the present invention.

Referring to Fig. 4, the artificial intelligence engine learning unit (105) of the image editing device (101) trains the artificial intelligence engine using a plurality of training image data collected from the image collection unit (103). From now on, the process of training the artificial intelligence engine will be described in detail with reference to Fig. 5.

Figure 5 is a flow chart of learning an artificial intelligence engine in an artificial intelligence engine learning unit (105) according to one embodiment of the present invention.

Referring to FIG. 5, the artificial intelligence engine learning unit (105) checks, in step 501, whether an image type for learning the artificial intelligence engine is input by the user. For example, the artificial intelligence engine learning unit (105) can receive an image type from the user through an input/output unit (not shown).

As a result of the verification, if the image type is input, the artificial intelligence engine learning unit (105) proceeds to step 503, otherwise, it repeatedly performs step 501.

In step 503, the artificial intelligence engine learning unit (105) selects a plurality of learning image data corresponding to the input image type from among a plurality of learning image data collected from the image collection unit (103).

In step 505, the artificial intelligence engine learning unit (105) trains the artificial intelligence engine by applying pre-specified scene learning parameters to the selected learning image data. At this time, the artificial intelligence engine learning unit (105) extracts at least one image data corresponding to at least one scene that satisfies the scene learning parameters from the selected learning image data, and generates the extracted image data as at least one scene learning image data. Then, the artificial intelligence engine learning unit (105) stores the generated scene learning image data in the scene learning database (107).

In step 507, the AI engine learning unit (105) trains the AI engine by applying pre-specified editing learning parameters to the selected learning image data. At this time, the AI engine learning unit (105) extracts at least one image data corresponding to at least one editing point that satisfies the editing learning parameters from the selected learning image data, and generates the extracted image data as at least one editing learning image data. Then, the AI engine learning unit (105) stores the generated editing learning image data in the editing learning database (109).

Returning to FIG. 4 again, the image synchronization unit (113) of the image editing device (101) synchronizes the reference points of a plurality of original image data input from the image input unit (111) at step 403.

For example, the video synchronization unit (113), as illustrated in FIG. 2, may perform sound synchronization by designating a pre-designated sound included in the original video data (123) as a reference point and setting specific frames (e.g., xxxxx year yy month zz day aa hour bb minute cc second dd frame) of the designated reference point as start vertical time codes (LTCs). In addition, the video synchronization unit (113) may perform video synchronization by designating frames of the original video data (e.g., xxxxx year yy month zz day aa hour bb minute cc second dd frame) as the same point in time as the start vertical time codes of the original video data (123).

At step 405, the scene analysis unit (115) of the image editing device (101) analyzes a plurality of scenes from a plurality of synchronized original image data using a pre-learned model.

In step 407, the editing analysis unit (117) of the video editing device (101) selects multiple scenes from among a plurality of scenes using an artificial intelligence program. Then, the editing analysis unit (117) automatically edits the selected scenes based on a plurality of editing learning video data to generate one editing video data (119) and editing information (121).

Through this process, one embodiment of the present invention can automatically edit image data using artificial intelligence in an image editing device, thereby editing image data in a short time and with little effort. And one embodiment of the present invention can automatically edit image data using artificial intelligence in an image editing device, thereby editing image data so that the result has a consistent quality.

Although the preferred embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific embodiments described above, and various modifications may be made by those skilled in the art without departing from the gist of the present invention as claimed in the claims. Furthermore, such modifications should not be individually understood from the technical idea or prospect of the present invention.

101: Video Editing Devices
103: Video Collection Unit
105: Artificial Intelligence Engine Learning Department
107; Scene Learning Database
109: Edit Learning Database
111: Video input
113: Video synchronization unit
115: Scene Analysis Department
117; Editing Analysis Department
119: Editing video data
121: Editing information
123: Multiple original video data
125: A large number of training image data

Claims (10)

An image collection unit that collects and stores a large number of image data for learning;
An image input unit that receives and stores a plurality of original image data;
A scene learning database that stores a number of scene learning video data by video type;
An editing learning database storing a plurality of editing learning video data by the above video type;
An artificial intelligence engine learning unit that selects a plurality of learning image data corresponding to a specific image type input by a user from among the above learning image data, and trains an artificial intelligence engine using the selected learning image data and pre-specified scene learning parameters and pre-specified editing learning parameters;
A video synchronization unit that synchronizes the reference point of the original video data based on a specific sound designated in advance;
A scene analysis unit that analyzes a plurality of scenes from the synchronized original image data using the scene learning image data; and
An editing analysis unit is included that applies the above-mentioned editing learning video data to an artificial intelligence program using the above-mentioned artificial intelligence engine to select a plurality of scenes from the analyzed scenes, and edits the selected scenes to generate edited video data.

The above artificial intelligence engine learning unit is,
The scene learning parameter is applied to the selected learning image data to train the artificial intelligence engine, and the scene learning image data satisfying the scene learning parameter is extracted from the selected learning image data and stored in the scene learning database.
The artificial intelligence engine is trained by applying the editing learning parameters to the selected learning image data, and the editing learning image data satisfying the editing learning parameters are extracted from the selected learning image data and stored in the editing learning database.
The above scene learning parameters include at least one of the number of speakers, object movement, screen composition, camera movement, and audio syllables.

The above editing learning parameters include at least one of scene transition, audio transition, dialogue recognition, action recognition, and background music position.
A device that automatically edits video data using artificial intelligence in a video editing device, wherein each of the above learning video data is of the same length and pre-edited according to video type.
delete delete delete delete The video collection unit collects and stores a large number of video data for learning.
The process of the artificial intelligence engine learning unit selecting multiple learning image data corresponding to a specific image type input by the user from among the learning image data;
The above artificial intelligence engine learning unit is a process for learning the artificial intelligence engine using the selected learning image data, pre-designated scene learning parameters, and pre-designated editing learning parameters.
The video input unit is a process of receiving and storing a large number of original video data.
The process of synchronizing the reference point of the original video data based on a specific sound designated in advance,
A process in which a scene analysis unit analyzes multiple scenes from the synchronized original image data using the scene learning image data.
The editing analysis department applies the editing learning video data to the artificial intelligence program using the artificial intelligence engine to select multiple scenes from among the analyzed scenes.
The above editing analysis unit includes a process of editing the selected scenes to generate edited video data.

The process of training the above artificial intelligence engine is:
A process of training the artificial intelligence engine by applying the scene learning parameters to the selected learning image data.
A process of extracting scene learning image data satisfying the scene learning parameters from the selected learning image data and storing them in a scene learning database.
A process of training the artificial intelligence engine by applying the above-mentioned editing learning parameters to the above-mentioned selected learning image data, and
A process of extracting the edited learning image data satisfying the edited learning parameters from among the selected learning image data and storing them in an edited learning database,

The above scene learning parameters include at least one of the number of speakers, object movement, screen composition, camera movement, and audio syllables.
The above editing learning parameters include at least one of scene transition, audio transition, dialogue recognition, action recognition, and background music position.
A method for automatically editing video data using artificial intelligence in a video editing device, wherein each of the above learning video data has the same length for each video type and is pre-edited.
delete delete delete delete

Images