Zero-sample image retrieval method and device based on hash coding and graph attention mechanism
A technology of hash coding and sample images, applied in the information field to promote knowledge transfer, avoid overfitting learning, and improve generalization ability
Active Publication Date: 2021-01-08
INST OF INFORMATION ENG CHINESE ACAD OF SCI
View PDF10 Cites 4 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0008] Aiming at the deficiencies of existing zero-sample hashing methods, the present invention proposes a zero-sample image retrieval method and device based on hash coding and graph attention mechanism
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
[0054] In order to make the above objects, features and advantages of the present invention more comprehensible, the present invention will be further described in detail below through specific embodiments and accompanying drawings.
[0055] The invention proposes a zero-sample image retrieval method based on hash coding and graph attention mechanism. This method provides an end-to-end neural network architecture, mainly composed of hash network, relation network and loss module. In the hash network, the image is obtained through a deep convolutional neural network to obtain image features, and then converted into a hash code through a fully connected layer. In order to fully mine and utilize the similarity relationship between categories, the relational network enhances the migration ability of the hash code by fusing the similarity relationship into the generation process of the hash code. First, the semantic similarity map is constructed by using the semantic vector of the...
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 relates to a zero-sample image retrieval method and device based on hash coding and a graph attention mechanism. The method comprises the following steps: constructing a hash network anda relationship network; training the hash network and the relational network based on the classification loss of the soft margin; inputting each image in the database into the trained hash network toobtain a corresponding image hash code; and inputting the to-be-queried image into the trained hash network to generate a hash code, calculating the distance between the hash code and the hash code of each image in the database, and returning a query result meeting the requirement according to the distance. According to the method and device, semantic and visual information can be considered at the same time, the similarity relation between classes is fully mined, knowledge migration is better achieved, hash learning is conducted based on classification losses of soft margins, overfitting learning of visible classes can be avoided to a certain extent, the generalization ability of the model for unseen classes is improved, and thus, the zero-sample image retrieval effect is improved.
Description
technical field [0001] The invention belongs to the field of information technology, and in particular relates to a zero-sample image retrieval method and device based on hash coding and graph attention mechanism. Background technique [0002] With the rapid development of the information age, image data shows an explosive growth trend, and the demand for efficient retrieval in massive images is increasing day by day. In real life, new categories are constantly appearing. How to adapt the model to the retrieval of new categories, that is, zero-shot image retrieval, has become a hot issue. This work has important practical application value in many fields such as intelligent monitoring, precision medicine, and e-commerce. In large-scale scenarios, zero-shot hashing methods can effectively improve computation and storage efficiency, and are widely used in zero-shot image retrieval tasks. Existing zero-shot hashing methods mainly include two stages, one is the extraction of i...
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 Applications(China)
IPC IPC(8): G06F16/532G06F16/55G06F16/583G06K9/62G06N3/04G06N3/08
CPCG06F16/532G06F16/55G06F16/583G06N3/084G06N3/045G06F18/22G06F18/241G06F18/25Y02D10/00
Inventor 吴大衍黄梅雪李波王伟平
Owner INST OF INFORMATION ENG CHINESE ACAD OF SCI