Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Video data processing method and device and electronic system

A technology for video data and processing methods, applied in character and pattern recognition, instruments, computer components, etc., can solve problems such as waste of computing power, difficulty in heat dissipation, and difficulty in realizing real-time calculations, and reduce data bandwidth and power consumption. The effect of calculation volume

Pending Publication Date: 2020-04-03
MEGVII BEIJINGTECH CO LTD
View PDF15 Cites 6 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Therefore, if each frame of the video is input into the neural network for calculation, it means that there will be a great waste of computing power and occupy a large data bandwidth, which will cause problems such as increased power consumption, difficulty in heat dissipation, and difficulty in real-time calculation.

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
  • Video data processing method and device and electronic system
  • Video data processing method and device and electronic system
  • Video data processing method and device and electronic system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0032] First, refer to figure 1 An example electronic system 100 for implementing the image-based method, device and electronic system for counting target objects according to an embodiment of the present invention will be described.

[0033] like figure 1 Shown is a schematic structural diagram of an electronic system, an electronic system 100 includes one or more processing devices 102, one or more storage devices 104, an input device 106, an output device 108, and one or more video capture devices 110, these components The interconnections are via bus system 112 and / or other forms of connection mechanisms (not shown). It should be noted that figure 1 The components and structures of the electronic system 100 shown are exemplary rather than limiting, and the electronic system may also have other components and structures as required.

[0034]The processing device 102 may be a smart terminal, or a device including a central processing unit (CPU) or other forms of processin...

Embodiment 2

[0041] This embodiment provides a method for processing video data, which improves the traditional task model for convolution calculation, such as figure 2 A flow chart of a processing method for video data shown, the processing method for video data includes the following steps:

[0042] Step S202, dividing the target video into key frames and normal frames.

[0043] The target video refers to the video used for convolution calculation. The format of the target video is not limited, as long as different video frames can be extracted from the target video. The target video includes several images, each of which is a video frame, and these video frames are divided into key frames and normal frames.

[0044] First, the initial video frame of the target video is divided into key frames, and then the remaining key frames are determined. Each key frame must be different from the previous key frame, that is, each key frame needs to continue the previous key frame with a differenc...

Embodiment 3

[0056] This embodiment provides another method for processing video data. This method improves the traditional task model for performing convolution calculations. The embodiment focuses on describing the specific process of performing convolution on the L+1th convolutional layer. In this embodiment, the task model includes multiple convolutional layers, and the video data is convoluted through the convolutional layers, such as image 3 Shown is a flow chart for convolution of the L+1th layer convolutional layer, the method for convolution of the L+1th layer convolutional layer includes the following steps:

[0057] Step S302, obtaining input information; the input information includes: the key features corresponding to the key frames output by the L-th convolutional layer and the common features corresponding to the ordinary frames; wherein, the key features in the input information of the first convolutional layer are the key Frames, common features are common frames.

[005...

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 provides a video data processing method and device and an electronic system. The video data processing method comprises the following steps: dividing a target video into a key frame anda common frame; for the key frame, inputting the key frame into the task model, and outputting a calculation result of the key frame; for the common frame, obtaining a mask between the common frame and the reference key frame, wherein the mask has a foreground motion area and a background static area, and the reference key frame is the key frame closest to the common frame; determining a feature region corresponding to the common frame according to the foreground motion region of the mask; and obtaining a calculation result of the common frame according to the reference key frame, the featureregion and the task model. According to the video data processing method, common features are not completely input into the task model, and only the feature region is input into the task model, so that the calculation amount of the neural network can be reduced, and the data bandwidth and the power consumption can be reduced, and real-time calculation can be realized on the basis of keeping the convolution calculation effect not to be reduced.

Description

technical field [0001] The present invention relates to the technical field of video data processing, in particular to a video data processing method, device and electronic system. Background technique [0002] Video data is a very common data form in IoT (Internet of Things, Internet of Things) scenarios. At present, tasks such as object detection, semantic segmentation, and video noise reduction are performed on video data based on neural networks. Usually, the video is regarded as a picture set, and each frame of the video is processed separately. [0003] In related technologies, each frame of a video is usually regarded as a picture, and all these pictures are input into a convolutional neural network for calculation. For high-resolution video, the number of frames is high, which requires a huge amount of calculation. Moreover, video data has great temporal and spatial redundancy. In the case that the camera does not move, several frames within a period of time may h...

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
IPC IPC(8): G06K9/62G06K9/00
CPCG06V20/46G06F18/211
Inventor 衡稳周舒畅
Owner MEGVII BEIJINGTECH CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products