A Video Retrieval and Matching Method Based on Dynamic Programming
A dynamic programming and matching method technology, applied in the field of video retrieval, can solve the problems that the video retrieval scheme is not suitable for large-scale video retrieval, etc., and achieve the effect of improving retrieval recall, accuracy and confidence
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment
[0068] A kind of video retrieval matching method based on dynamic programming in this embodiment, such as figure 1 shown, including the following steps:
[0069] S1: The video is evenly drawn at a fixed frequency, and each frame is extracted with a pre-trained neural network to extract the feature vector.
[0070] In this embodiment, for existing stock or newly uploaded videos, frames are drawn evenly at 1 frame per second. Each frame of picture extracts feature vectors through a large-scale pre-trained CNN network.
[0071] The CNN network here can be a picture feature extraction network pre-trained on any big data, such as resnet-18 trained on imagenet, or the ViT network released by Google. Its input is an image, and the output is a one-dimensional feature vector.
[0072] For example, the video has 10 seconds, one frame per second, and each frame is extracted to a 1000-dimensional vector through the CNN network, and the frame feature vector sequence is a 10x1000-dimensi...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More 


