Method, device, medium and equipment for training and use of floating object recognition model

Abstract

The application discloses a method, device, medium and equipment for training and using a floating object recognition model, belonging to the technical field of image processing. The method includes: acquiring a training data set, the training data set includes multiple groups of training samples, and each group of training samples includes an image and a corresponding mask map; using the spatial path in the bilateral high-resolution segmentation model to perform the training samples. process to obtain the first feature map; use the context path in the bilateral high-resolution segmentation model to process the training samples to obtain four second feature maps; perform feature fusion and segmentation according to the four second feature maps and the first feature map After decoding, four branch loss functions are obtained; a bilateral high-resolution segmentation model is trained according to the four branch loss functions, and a floating object recognition model is obtained. The present application can not only improve the detection efficiency and accuracy of floating objects, but also improve the accuracy of the floating object recognition model through the auxiliary supervision training of multiple branches.

Description

technical field [0001] The present application relates to the technical field of image processing, in particular to a method, device, medium and equipment for training and using a floating object recognition model. Background technique [0002] River floating objects will not only cause river pollution, but also affect the viewing of the river. River floating objects are mainly garbage (such as plastic bags, plastic bottles, etc.), duckweed, aquatic plants, etc. Therefore, the most effective way to control river pollution is to deal with floating objects in a timely manner to avoid the expansion of the polluted area and reduce the impact on microorganisms in the river. harm. [0003] When artificial river inspection is used to control river pollution, the artificial inspection of the river is inevitably affected by factors such as the length of the river, visual blind spots, and human eyes have certain limitations, resulting in some floating objects that are easy to be igno...

AI Technical Summary

Problems solved by technology

[0004] This application provides a method, device, medium and equipment for training and using a floating object recognition model, which are used to solve the problem that floating objects are easily overlooked when manually inspecting river channels

Images