Image hash code training model algorithm and classification learning method based on binary weight

Abstract

The invention discloses an image hash code training model algorithm and a classification learning method based on binary weights. The model algorithm steps include: selecting a loss function, determining an objective equation, and performing binary encoding on classifiers and training image features; Code for unified learning, update binary code, optimize loss function; derive hash code training model. Also disclosed is a classification learning method using a hash code image training model based on binary weights. The steps include: the image to be searched is obtained through the hash code training model based on binary weights to obtain the hash code, and the hash code and the hash code are obtained. The Hamming distance between the binary codes of the classifiers; find the minimum Hamming distance in the Hamming distance, and obtain the corresponding classifier, which is the category of the image to be searched. The invention can classify images in multiple image categories and high-latitude scenes, improve the performance of algorithms on large-scale data sets, and is accurate, efficient, fast and low in memory consumption.

Description

technical field

[0001] The invention belongs to the field of image classification methods, and in particular relates to a binary weight-based image hash code training model and a classification learning method. Background technique

[0002] In recent years, due to the explosive growth of the number of digital images and the substantial improvement of image quality, the problem of large-scale visual recognition has attracted great research enthusiasm from academia and industry. The classification problem of images with tens of thousands of categories usually uses conventional classifiers for heavy calculations, such as k-nearest neighbors or k-NN, and support vector machines or SVM. In multi-class image recognition problems, a large number of classifiers produces huge computational and memory overheads, and a large number of classifiers will lead to a surge in complexity during the model training and deployment stages. Imagine that there are C categories, and each category h...

Examples

Embodiment 1

[0067] An image hash code training model algorithm based on binary weights, the algorithm includes the following steps:

[0068] Step 1.1. Select the exponential loss function as the loss function used, and the formula of the objective equation is:

[0069]

[0070]

[0071] Let the generated image binary code be b i is the original training dataset middle x i For the corresponding r-bit binary code, a linear hash equation is set as:

[0072] b=sgn(P T x)

[0073] here P is the image hash transpose matrix; T is the transpose symbol; d is the dimension of image x; r is the hash code length;

[0074] The binary code of the classifier is w.

[0075] Step 1.2. Perform unified learning on the classifier obtained in step 1.1 and the binary code of the training image features, update the hash code of the training image features and the binary code of the classifier, optimize the objective equation of the loss function selected in step 1.1, and get optimized After th...

Embodiment 2

[0106] On the basis of Embodiment 1, different loss functions are selected, and the steps of a hash code image training model algorithm based on binary weights are as follows:

[0107] Step 1.1. Select a simple linear loss function as the loss function used, and the formula of the objective equation is:

[0108]

[0109]

[0110] Let the generated image binary code be b i is the original training dataset middle x i For the corresponding r-bit binary code, a linear hash equation is set as:

[0111] b=sgn(P T x)

[0112] here P is the image hash transpose matrix; T is the transpose symbol; d is the dimension of image x; r is the hash code length;

[0113] The binary code of the classifier is w.

[0114] Step 1.2: Perform unified learning on the classifier obtained in step 1.1 and the binary code of the training image feature, update the hash code of the training image feature and the binary code of the classifier, optimize the objective equation of the loss func...

technical field

[0001] The invention belongs to the field of image classification methods, and in particular relates to a binary weight-based image hash code training model and a classification learning method. Background technique

[0002] In recent years, due to the explosive growth of the number of digital images and the substantial improvement of image quality, the problem of large-scale visual recognition has attracted great research enthusiasm from academia and industry. The classification problem of images with tens of thousands of categories usually uses conventional classifiers for heavy calculations, such as k-nearest neighbors or k-NN, and support vector machines or SVM. In multi-class image recognition problems, a large number of classifiers produces huge computational and memory overheads, and a large number of classifiers will lead to a surge in complexity during the model training and deployment stages. Imagine that there are C categories, and each category h...

Examples

Embodiment 1

[0067] An image hash code training model algorithm based on binary weights, the algorithm includes the following steps:

[0068] Step 1.1. Select the exponential loss function as the loss function used, and the formula of the objective equation is:

[0069]

[0070]

[0071] Let the generated image binary code be b i is the original training dataset middle x i For the corresponding r-bit binary code, a linear hash equation is set as:

[0072] b=sgn(P T x)

[0073] here P is the image hash transpose matrix; T is the transpose symbol; d is the dimension of image x; r is the hash code length;

[0074] The binary code of the classifier is w.

[0075] Step 1.2. Perform unified learning on the classifier obtained in step 1.1 and the binary code of the training image features, update the hash code of the training image features and the binary code of the classifier, optimize the objective equation of the loss function selected in step 1.1, and get optimized After th...

Embodiment 2

[0106] On the basis of Embodiment 1, different loss functions are selected, and the steps of a hash code image training model algorithm based on binary weights are as follows:

[0107] Step 1.1. Select a simple linear loss function as the loss function used, and the formula of the objective equation is:

[0108]

[0109]

[0110] Let the generated image binary code be b i is the original training dataset middle x i For the corresponding r-bit binary code, a linear hash equation is set as:

[0111] b=sgn(P T x)

[0112] here P is the image hash transpose matrix; T is the transpose symbol; d is the dimension of image x; r is the hash code length;

[0113] The binary code of the classifier is w.

[0114] Step 1.2: Perform unified learning on the classifier obtained in step 1.1 and the binary code of the training image feature, update the hash code of the training image feature and the binary code of the classifier, optimize the objective equation of the loss func...

Images