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

DHT network-based R-CNN network optimization method and a storage medium

An R-CNN and network optimization technology, applied in the object detection field of the R-CNN network, can solve problems such as large disk space occupation, increased search difficulty, and inability to classify data, so as to reduce memory occupation and shorten search difficulty , the effect of less resource consumption

Active Publication Date: 2020-05-08
FUJIAN TQ DIGITAL
View PDF6 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] In the above operation, after extracting about 2000 candidate frames, the similarity of each data in the candidate frames needs to be repeatedly calculated and merged. To save, it takes up a lot of disk space, and does not classify the data, which increases the difficulty of searching

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
  • DHT network-based R-CNN network optimization method and a storage medium
  • DHT network-based R-CNN network optimization method and a storage medium
  • DHT network-based R-CNN network optimization method and a storage medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0076] Please refer to Figure 2 to Figure 6 , this embodiment provides an R-CNN network optimization method based on the DHT network. By optimizing the algorithm and the data storage method, efficient calculations can be achieved with less space occupation, and at the same time, the search difficulty can be reduced.

[0077] see image 3 , methods include:

[0078] S0: The picture is passed to the R-CNN network to generate a preset number of candidate boxes.

[0079] After input, the system will automatically generate about 2000 candidate boxes;

[0080] S1: Extract the feature value set corresponding to each candidate frame; the feature value set includes a set of feature values ​​such as color space, color distance, texture distance, size, and shape overlap.

[0081] Preferably, specific marks are set for the above feature values, for example, the color space is marked as cs, the color distance is marked as cd, the texture distance is marked as td, the size is marked as ...

Embodiment 2

[0104] see Figure 5 to Figure 8 , this embodiment provides a specific application scenario corresponding to Embodiment 1:

[0105] The R-CNN network optimization scheme based on the DHT network distribution method optimizes the selective search algorithm and data storage in the object detection method, and completes object detection faster and takes less resources.

[0106] First, after a picture is passed in, a DHT network is established, and 5 storage blocks are set up, which are used to store color space cs, color distance cd, texture distance td, size si, and shape overlap notation os, these five types of eigenvalues, The mean values ​​stored by the nodes of each storage block are recorded as avg_cs, avg_cd, avg_td, avg_td, avg_si and avg_os.

[0107] Perform a selective search algorithm such as figure 1 As shown, the first two texture distances of the upper left corner of the storage block storing the eigenvalue color space td in the DHT network are calculated to obtai...

Embodiment 3

[0114] This embodiment corresponds to Embodiment 1 and Embodiment 2, and provides a computer-readable storage medium on which a computer program is stored. After the program is executed by a processor, it can realize the above-mentioned embodiment 1 or embodiment 2. The steps included in the R-CNN network optimization method based on the DHT network. The specific steps will not be repeated here, please refer to the description of Embodiment 1 or Embodiment 2 for details.

[0115] In summary, the DHT network-based R-CNN network optimization method and storage medium provided by the present invention can significantly improve computing power and efficiency, and obtain calculation results more quickly and efficiently; at the same time, it can reduce memory usage and optimize memory usage. and system performance; further, it can also reduce the difficulty of searching.

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 DHT network-based R-CNN network optimization method and a storage medium. The method comprises the following steps that: feature value sets corresponding to various candidateboxes are extracted; feature values in the feature value sets are stored into different storage blocks of a DHT network in a classified manner, the storage areas of the feature values in the storageblocks are corresponding to the layouts of the candidate boxes of the feature values in a picture; whether the mean values of the front and rear groups of feature values in each storage block are equal or not is sequentially calculated by using a non-equivalent carry mean value algorithm, the storage areas of the two groups of feature values with the equal mean value are merged, and the mean values is taken as the node ID of the merged storage areas, and two adjacent feature values are adopted as a group of feature values; the node IDs of the storage blocks of the DHT network are normalized, and the normalized node IDs are transmitted to an R-CNN network. With the method adopted, computing power and efficiency can be remarkably improved; a computing result can be obtained more quickly andefficiently; memory occupation can be reduced, and memory and system performance can be optimized; and searching difficulty can be reduced.

Description

technical field [0001] The invention relates to the object detection field of R-CNN network, in particular to a DHT network-based R-CNN network optimization method and a storage medium. Background technique [0002] R-CNN is a region-based convolutional neural network. The current implementation method is to input an image; use a selective search algorithm to extract about 2000 candidate boxes in the image, and scale the candidate boxes to a fixed size; input the normalized candidate boxes into the CNN network to extract features; for each The CNN features extracted from each candidate frame are identified by SVM classification, the position and size of the frame are fine-tuned by linear regression, and a frame regressor is trained for each category separately. The whole process is the initial region-based convolutional neural network. (R-CNN) implementation. [0003] For the optimization algorithm of R-CNN, there are currently existing optimization methods: whether it is ...

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): G06N3/04G06N3/063G06N3/08
CPCG06N3/063G06N3/08G06N3/045
Inventor 刘德建于恩涛陈琦林小云张小华林琛
Owner FUJIAN TQ DIGITAL