Incremental small sample target detection method and system based on weight generation

Abstract

The invention belongs to the field of computer vision, and specifically relates to an incremental small-sample target detection method and system based on weight generation, aiming to solve the problem that existing target detectors lack the ability of small-sample rapid learning and incremental learning, and the label data The dependence is strong, and the detector does not have open problems. The present invention includes: performing supervised training of detectors through basic category data; obtaining weights of scale perception and centrality perception of basic category target detectors to generate basic category responses; combining basic category responses to generate new category weights; combining new category data for basic Fine-tuning training of class object detectors; Incremental few-shot object detection through obtained base class and new class object detectors. The invention combines the perception of scale and centrality, the regional features are more representative, the target positioning is more accurate, the model can obtain better overall performance in incremental learning, and the detection efficiency, accuracy and precision are high.

Description

technical field [0001] The invention belongs to the field of computer vision, and in particular relates to an incremental small sample target detection method and system based on weight generation. Background technique [0002] As an essential step in many computer vision tasks, object detection has attracted extensive attention from academia and industry in recent years. With the help of deep neural networks, the field of object detection has made great progress. However, deep neural networks are highly dependent on huge training data and manual labeling that consumes a lot of manpower, and in many practical situations, it is not easy for people to get massive labels. At the same time, when faced with a data stream that may increase new detection requirements over time, most object detectors based on deep neural networks lack the ability of fast learning and incremental learning with small samples. [0003] Recently, several studies have proposed incremental / few-shot lear...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the above-mentioned problems in the prior art, that is, the existing target detector lacks the ability of small-sample rapid learning and incremental learning, has a strong dependence on label data, and the detector does not have annotation efficiency and openness, the present invention provides An incremental small-sample object detection method based on weight generation is proposed, which includes:

Images