Few-sample target detection method based on meta-feature and weight adjustment and network model

Abstract

The invention discloses a few-sample target detection method based on meta-feature and weight adjustment and a network model. The method comprises the following steps: S1, constructing a detection network model and preprocessing an image; s2, extracting meta-features and weight vectors of the base class images; s3, combining the extracted meta-features and weight vectors to obtain a multi-dimensional feature map, and inputting the multi-dimensional feature map into a classification regression module to calculate a loss function; s4, adjusting network parameters according to the loss function and the gradient descent, and realizing training of a detection network model by the base class image; s5, extracting meta-features and weight vectors of the base class and new class joint images; s6,repeating the step S3 and the step S4, and training of the new class and base class combined image on the detection network model is completed; and S7, detecting the test image by using the trained detection network model. According to training of the detection network model, meta-features are extracted by using samples of a large amount of data, and fine adjustment is performed by means of few sample data, so that the target detection accuracy of a small amount of marked samples is improved.

Description

technical field [0001] The invention belongs to the technical field of computer vision detection, and mainly relates to deep learning target detection. Specifically, it proposes a few-sample target detection method and network model based on meta-features and weight adjustments, which can be used to realize target detection with only a small number of samples. Classify and locate. Background technique [0002] At present, the target detection task in the field of computer vision has been widely used in the fields of industrial production and intelligent monitoring. Target detection is extended from image classification, which mainly includes identifying the target contained in the image and marking the position of the target. In the past, due to the limitations of computer processing speed and memory, researchers generally used traditional non-convolutional neural network detection methods to detect targets, but with the rapid development of computer processing speed and me...

AI Technical Summary

Problems solved by technology

In terms of speed and accuracy, the two detection methods have their own advantages. Generally speaking, the detection accuracy of two-stage target detection is higher than that of single-stage target detection. The speed is better than the two-stage target detection, but the target detection task in the two detection methods is heavily dependent on a large number of labeled data sets for training, but in practical applications, some objects have little data or are difficult to detect. Obtain

When there is a lack of labeled data, severe overfitting and very poor generalization capabilities will occur, resulting in low detection accuracy or no detection at all.

Images