Small sample target detection method and system based on category semantic feature reweighting

Abstract

The invention discloses a small sample target detection method and system based on category semantic feature reweighting. The relevance between a base category and a new category is calculated by using the semantic information of a category label, then the meta-feature of the base category is transmitted to the meta-feature of the new category by using a graph convolutional network according to the relevance degree between the base category and the new category. Good new class meta features can be learned under the condition that only a small amount of new class data exists. A support set and a query sample are constructed according to a base class and a new class; A category semantic graph is constructed according to the base category and the new category; A category semantic embedding module is constructed. the whole network is trained in a two-stage training mode, the network comprises a feature extractor, a meta-learning device, a category semantic embedding module and a detection layer, and the effectiveness of the method is proved through a contrast experiment on PASCAL VOC.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a small-sample target detection method and system based on category semantic feature reweighting. Background technique [0002] In recent years, due to the rapid development of computer hardware and the emergence of large-scale labeled datasets such as ImageNet, artificial intelligence has developed tremendously, making it surpass humans in many fields. However, there are still many problems and challenges for artificial intelligence to use a small number of samples to learn new knowledge and use it to solve practical problems like humans. [0003] As a data-driven technology, deep learning generally relies on a large amount of training data to enable the model to achieve good results. However, in real-world scenarios, collecting a large amount of data and labeling it accordingly requires a lot of energy and time for professionals, even in some special applic...

AI Technical Summary

Problems solved by technology

[0007] The technical problem to be solved by the present invention is to provide a small-sample object detection method and system based on category semantic feature reweighting to solve the problems in the existing small-sample object detection method based on meta-learning. , when learning the meta-features of the base class and the new class, aiming at the problem that the relationship between the base class and the new class category is not fully utilized, resulting in poor expressive ability of the meta-features of the new class

Images