Unlock instant, AI-driven research and patent intelligence for your innovation.

Convolutional neural network similar video retrieval method based on attention mechanism

A convolutional neural network and attention technology, applied in the field of similar video retrieval, can solve problems such as difficult to retrieve, not very accurate, and difficult to use, so as to improve retrieval efficiency, reduce response, and improve accuracy. Effect

Active Publication Date: 2022-06-03
BEIJING SCISTOR TECH
View PDF6 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003]At present, the similar video detection method is mainly based on two aspects: 1. The overall retrieval of the video. It is a feature value, but it often does not perform very well in terms of accuracy. Once the video has been modified, intercepted, spliced, etc., it is difficult to be retrieved
2. Extract the video into frame-by-frame images for retrieval. The accuracy of this method is often very good, but the speed will be very slow. When a video is very long, it will take a long time to return the result, so it is very slow Difficult to use in specific applications

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Convolutional neural network similar video retrieval method based on attention mechanism
  • Convolutional neural network similar video retrieval method based on attention mechanism
  • Convolutional neural network similar video retrieval method based on attention mechanism

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0021] The present invention will be further described in detail below in conjunction with the accompanying drawings.

[0022] like figure 1 As shown, the convolutional neural network similar video retrieval method based on the attention mechanism of the present invention, such as figure 1 shown, the specific steps are as follows:

[0023] Step 1: Retrieve the keyframe extraction of the video.

[0024] Because there is a large amount of repeated data between frames, the block structure idea is used instead of the traditional continuous structure idea, and a frame of picture extracted from the block is called a key frame, so it can greatly reduce the number of video frame extraction pictures. The speed of retrieval of the entire video is improved.

[0025] like figure 2 As shown, in the present invention, for the input video of the video, the first video picture that is not a solid color (one frame of picture is the same pixel value) is used as the first key frame of the s...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a convolutional neural network similar video retrieval method based on an attention mechanism. The method comprises the following steps: extracting key frames of a retrieved video, and replacing a continuous structure thought with a block structure thought. Video key frame image processing: introducing a pure color removal algorithm and enhancing overall and local features of the image; and key frame feature extraction: extracting key frame features by using improved ResNet-50. And video key frame feature similarity retrieval is carried out, and Fiss retrieval is introduced. And post-processing an inter-frame result, and introducing a straightening machine and a Softmax mechanism. According to the method, the problems of time and precision of large-scale similar video retrieval are mainly solved, the retrieval duration is effectively shortened while the precision is not reduced, and the video retrieval performance is greatly improved.

Description

technical field [0001] The invention belongs to the technical field of similar video retrieval, and relates to a convolutional neural network based on an attention mechanism, which realizes retrieval and identification of similar videos, and adopts corresponding means to improve the accuracy and speed. Background technique [0002] With the era of big data, the Internet is full of massive amounts of data. As an important part of it, video data is affecting people's daily life from all aspects. How to realize similar video retrieval in massive video data has a wide range of business applications, such as similar video de-duplication, similar video retrieval and so on. However, due to the birth of various video editing software, this work has become extremely difficult. The reason is that once the video has been edited and modified, it is difficult to match the original video using traditional methods. Generally, these modified videos generally have the following characteri...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06F16/783G06F16/78G06F16/71G06V20/40G06V10/74G06V10/80G06V10/82G06K9/62G06N3/04G06N3/08
CPCG06F16/7844G06F16/7867G06F16/71G06N3/08G06N3/045G06F18/22G06F18/253Y02D10/00
Inventor 谢铭吴林涛董建武索帅郑博文王立刚蔡荣华胡小勇
Owner BEIJING SCISTOR TECH