Weather-adaptive target detection network model and method

Abstract

The invention discloses a weather-adaptive target detection network model and method, and the method comprises the steps: collecting sunny day video data and rainy day video data through a visual collection device, and writing a script to process a video into a picture; marking the sunny day image data, taking a marked sunny day data set as a source domain data set, and taking an unmarked rainy day data set as a target domain data set; putting the source domain data set and the target domain data set into a weather-adaptive target detection network model for training to obtain a weight parameter file of the model; loading the weight parameter file to obtain a detection network model; and extracting a picture from the video stream according to a set frame number, inputting the picture into the detection network model for prediction, and displaying a prediction result. The detection precision is guaranteed and even improved under the weather influence condition, and meanwhile the problem that the workload of manual data labeling is large is solved by the method.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a domain-based adaptive target detection network model and a detection method. Background technique [0002] There are two important problems in the current target detection network such as yolov5, faster-rcnn, FCOS, etc. First of all, these detection networks rely on a large amount of labeled training data. Accurately labeling all the training data usually costs a lot of money, and the training data for some scenarios is not easy to obtain. Secondly, these networks have the problem of insufficient model generalization ability. For example, when the weather changes, such as sunny and rainy days, the detection accuracy of previous target detection algorithms will decrease. Contents of the invention [0003] The present invention mainly solves the problem that the existing practical target detection model is greatly affected by the weather, and the model trai...

AI Technical Summary

Problems solved by technology

[0003] The present invention mainly solves the problem that the existing practical target detection model is greatly affected by the weather, and the model trained in the normal weather data set may not perform well in rainy and foggy weather, and provides a weather-adaptive target detection The network model and detection method ensure or even improve the detection accuracy under the influence of weather. At the same time, this method also solves the problem of heavy workload of manual labeling data

Images