Training method and device of detection segmentation model, and target detection method and device

Abstract

The invention discloses a training method for a detection segmentation model, and the method comprises the steps: in a first stage, training a to-be-trained detection segmentation model through employing pixel-level labeling sample data, and obtaining a detection segmentation model with a first model parameter; and, in a second stage, during each training, randomly selecting one piece of image-level category annotation sample data and one piece of pixel-level annotation sample data, respectively inputting the image-level category annotation sample data and the pixel-level annotation sample data into the current detection and segmentation model, training the current detection and segmentation model, repeatedly iterating the process of the second stage until all training is finished to obtain a detection and segmentation model with a second model parameter, and taking the current detection segmentation model as a trained detection segmentation model. According to the invention, the segmentation accuracy of the detection segmentation model is improved, and the quantity demand of the pixel-level marked sample data is reduced.

Description

technical field [0001] The invention relates to the field of machine learning, in particular to a training method for a detection and segmentation model. Background technique [0002] With the development of artificial intelligence (AI) technology, the machine learning model is the basis for realizing artificial intelligence technology. The accuracy of the machine learning model for task processing depends on the training of the machine learning model using the training set containing sample data. [0003] The detection and segmentation model can be used for target detection and to identify different instances of the same category of targets, for example, to identify different forms of sheep in the image (different forms can be understood as different instances), so that the detection and segmentation model can also be used for the same category Target for target segmentation. [0004] However, the training of detection and segmentation models relies on large-scale pixel-le...

AI Technical Summary

Problems solved by technology

According to relevant statistical results, the average time for pixel-level labeling samples of a frame of image is about 1 minute. Obtaining large-scale and accurate pixel-level labeling sample data requires a lot of manpower and time costs.

Images