Intelligent video editing method and system
An intelligent video editing method including: receiving and decoding an image data stream from a source storage device to generate an image sequence; performing image extraction operations on the image sequence at intervals of a predetermined time to obtain a plurality of image shots; selecting a frame out of each aforementioned image shot as a candidate frame, and performing a comparison operation on any two consecutive aforementioned candidate frames to derive a group of aforementioned candidate frames belonging to a same scene, and selecting one aforementioned candidate frame from the group as a target frame; and performing an AI evaluation operation on each aforementioned target frame to classify and/or rate each aforementioned target frame.
1. An intelligent video editing method implemented by an information processing device, including:
receiving and decoding an image data stream from a source storage device to generate an image sequence consisting of frames;
performing image extraction operations on the image sequence at intervals of a predetermined time to obtain a plurality of image shots, where each of the image shots has n said frames, and n is an integer greater than or equal to 1;
selecting an i-th one of the n aforementioned frames in each of the image shots as a candidate frame, i being an integer greater than or equal to 1 and less than or equal to n, and performing a comparison operation on any two consecutive said candidate frames to derive a group of said candidate frames belonging to a same scene, and selecting one said candidate frame from the group as a target frame, wherein the comparison operation is used to compare a difference value of two moment vectors of any two consecutive said candidate frames with a threshold, and the two candidate frames will be attributed to said same scene when the difference value is less than the threshold, wherein the difference value can be a root mean square error or a cosine distance; and
using a first neural network to perform a classification evaluation operation on the target frames to attribute each of the target frames to one of a plurality of predetermined classifications; or using a second neural network to perform a score evaluation operation on the target frames to give each of the target frames a score, the score being a real number; or using a first neural network to perform a classification evaluation operation on the target frames to attribute each of the target frames to one of a plurality of predetermined classifications, and using a second neural network to perform a score evaluation operation on the target frames to give each of the target frames a score, the score being a real number; where a plurality of weighting coefficients in the first neural network are obtained by using plural data sets of (input data, output data) to perform a training procedure on the first neural network, the input data being provided by a plurality of the frames prepared in advance, and the output data being provided by a plurality of predetermined classifications, and a plurality of weighting coefficients in the second neural network are obtained by using plural data sets of (input data, output data) to perform a training procedure on the second neural network, the input data being provided by a plurality of the frames prepared in advance, and the output data being provided by a plurality of predetermined real numbers.
2. The intelligent video editing method as disclosed in claim 1 , further including: storing at least one said score, which is generated after using the second neural network to perform the score evaluation operation, into a memory unit.
3. The intelligent video editing method as disclosed in claim 2 , wherein when the intelligent video editing method utilizes the second neural network to perform the score evaluation operation, the information processing device further provides a user interface for the user to set the predetermined score and derive at least one said target frame having the score higher than the predetermined score from the memory unit.
4. The intelligent video editing method as disclosed in claim 3 , wherein the information processing device further includes a score error recording module for storing at least one erroneous result of the score evaluation operation, which is reported by the user through the user interface, into the memory unit, so as to be used to improve the second neural network.
5. The intelligent video editing method as disclosed in claim 1 , further including: storing at least one said score, which is generated after using the second neural network to perform the score evaluation operation and is higher than a predetermined score, into a memory unit.
6. The intelligent video editing method as disclosed in claim 1 , further including: sorting and storing at least one said score, which is generated after using the second neural network to perform the score evaluation operation, into a memory unit.
7. The intelligent video editing method as disclosed in claim 1 , wherein the source storage device is a cloud storage device or a local storage device of the information processing device.
8. The intelligent video editing method as disclosed in claim 1 , wherein when the intelligent video editing method utilizes the first neural network to perform the classification evaluation operation, the information processing device further provides a user interface for a user to select an said predetermined classification and/or an said predetermined score to get at least one desired said target frame from a memory unit.
9. The intelligent video editing method as disclosed in claim 8 , wherein the information processing device further includes a classification error recording module for storing at least one erroneous result of the classification evaluation calculation, which is reported by a user through the user interface, into the memory unit, so as to be used to improve the first neural network.
10. The intelligent video editing method as disclosed in claim 8 , wherein the information processing device further includes a preference tracking module for storing records of the user's selections of the predetermined classifications into the memory unit.
11. The intelligent video editing method as disclosed in claim 10 , wherein the information processing device further includes an aversion tracking module for storing records of aversion feedbacks of at least one of the predetermined classifications, which are reported by the user through the user interface, into the memory unit.
12. The intelligent video editing method as disclosed in claim 11 , wherein the information processing device further includes an advertisement providing module for providing the user with at least one advertising message by referencing the records of the user's selections stored in the preference tracking module and/or the records of aversion feedbacks stored in the aversion tracking module.
13. The intelligent video editing method as disclosed in claim 11 , wherein the information processing device further includes a filtering module to filter messages for the user by referencing the records of aversion feedbacks stored in the aversion tracking module.
14. An intelligent video editing system, which has the information processing device as disclosed in claim 1 to implement the intelligent video editing method.
15. The intelligent video editing system as disclosed in claim 14 , wherein the information processing device is selected from a group consisting of a desktop computer, a portable computer, a smart phone, and a network attached storage device.