Cross-domain target detection method based on regional full convolutional network and self-adaptation

Abstract

The invention discloses a cross-domain target detection method based on a regional full convolutional network and self-adaptation, and belongs to the technical field of computer vision. According to the method, a deep learning target detection technology is used, and the cross-domain robustness of target detection is improved by using an adaptive method in allusion to the problem of different distribution of data of a training domain and a test domain in target detection. The method comprises steps of firstly, constructing a regional full convolutional network model based on deep learning; designing two corresponding domain classifiers on the image level and the target level as adaptive components to reduce the difference of domain transformation, and adding consistency regularization to the domain classifiers; training the network in an end-to-end mode; and finally, removing adaptive components, and applying the network to a target detection task. By adopting the cross-domain target detection method designed by the invention, the average precision of target detection in various domain transformation scenes can be effectively improved.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a region-based full convolution network and an adaptive cross-domain target detection method. Background technique [0002] Object detection is a fundamental problem in computer vision, which aims to detect and recognize all target objects corresponding to a certain category in an image. Object detection can be traced back a long time, and there have been many classical and effective methods. Classical works usually define object detection as a sliding window classification problem. In computer vision, the rise of deep convolutional networks originated from object detection. Driven by the rapid development of deep convolutional networks, researchers have proposed many object detection algorithms based on convolutional neural networks, which have greatly improved the performance of object detection. Among the large number of methods that have been proposed, regional ful...

AI Technical Summary

Problems solved by technology

[0007] In order to solve the problem of cross-domain target detection, the present invention provides a cross-domain target detection method based on regional full convolution network and self-adaptation, using deep learning target detection technology, aiming at the problem of different distribution of data in the training domain and test domain in target detection , using an adaptive method to improve the cross-domain robustness of object detection

Images