Fine-grained Image Classification Method Based on Attention Transfer Mechanism
A classification method and attention technology, applied in the field of computer vision, can solve the problem of low accuracy of fine-grained image classification, and achieve the effect of strengthening feature extraction ability and improving accuracy.
Active Publication Date: 2022-03-22
XIDIAN UNIV
View PDF5 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0008] The purpose of the present invention is to address the deficiencies of the above-mentioned prior art, and propose a fine-grained image classification method based on an attention transfer mechanism, which is used to solve the technical problem of low accuracy of fine-grained image classification in the prior art
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0055] The present invention will be described in further detail below in conjunction with the accompanying drawings and specific embodiments.
[0056] refer to figure 1 , a fine-grained image classification method based on the attention transfer mechanism, including the following steps:
[0057] Step 1) Obtain training sample set and test sample set:
[0058] Step 1a) The present embodiment adopts the California bird database CUB-200-2011, including 11,788 natural images of 200 kinds of birds, such as black-footed albatross, yellow-billed cuckoo, egret sparrow, etc.;
[0059] Step 1b) Perform data enhancement on all natural bird images in the database, including random rotation of [-10,+10] on all natural bird images, and then randomly flip the images horizontally;
[0060] Step 1c) normalize the size of the enhanced bird natural image, and the normalized bird natural image pixel is 448×448;
[0061] Step 1d) Select 5994 images of all size-normalized bird natural images fo...
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
Login to View More Abstract
The invention proposes a fine-grained image classification method based on an attention transfer mechanism, which is used to improve the fine-grained image classification accuracy. The implementation steps are: obtain the training sample set and test sample set containing fine-grained images; construct the global perception network and attention transfer network model; use the training sample set to train the global perception network and attention transfer network; use the trained global A perception network and an attention transfer network classify the test sample set. The present invention extracts joint features of images by designing a global perception network and an attention transfer network, so that the network further extracts the semantic correlation between different discriminative areas on the basis of extracting global features and discriminative regional features, thereby strengthening the The feature extraction capability of the network effectively improves the accuracy of fine-grained image classification.
Description
technical field [0001] The invention belongs to the technical field of computer vision and relates to a fine-grained image classification method, in particular to a fine-grained image classification method based on an attention transfer mechanism, which can be used for fine classification tasks, such as bird classification and car classification. Background technique [0002] Image classification is a processing method that extracts the discriminative features of the image itself through a design-based or learning-based method, so that smart devices can automatically identify the category to which the image subject belongs. Image classification methods are widely used in various fields of society, such as face recognition, species recognition, etc. According to the granularity of the image classification target, image classification methods can generally be divided into two categories, namely general (coarse-grained) image classification and sub-category (fine-grained) image...
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
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06F16/55G06K9/62G06N3/08
CPCG06F16/55G06N3/08G06V10/44G06V10/454G06F18/24
Inventor 牛毅焦阳李甫石光明
Owner XIDIAN UNIV