Shadow detection method based on attention mechanism

Abstract

The invention discloses a shadow detection method based on an attention mechanism. The shadow detection method comprises the steps of obtaining a to-be-processed public shadow data set for a trainingsystem; obtaining a to-be-detected shadow image from camera equipment or a local hard disk; preprocessing the public shadow data set and obtaining a corresponding training set and a corresponding testset; preprocessing the to-be-detected shadow image; constructing and fusing all modules of a system core neural network to form a convolutional neural network based on an attention mechanism; calculating network prediction and label loss, and adjusting network parameters according to the loss; completely training the deep convolutional neural network and inputting a preprocessed to-be-detected shadow image into the deep convolutional neural network; and outputting a shadow detection result, and performing shadow attribute classification on each pixel to complete a shadow detection process ofthe custom data. According to the invention, the shadow feature extraction capability is improved, the semantic relevance is reduced, the generalization and high efficiency of the detection system areimproved, and the shadow detection result is more accurate.

Description

technical field [0001] The present invention proposes an effective method for detecting shadows by studying shadow models and attention models, combined with residual convolutional neural networks. The ability to extract shadow features is increased, the generalization and efficiency of the detection system are improved, and the shadow detection results are more accurate, which belongs to the field of shadow detection. Background technique [0002] In computer vision classification tasks such as image recognition or semantic segmentation, the algorithm needs to extract the features in the image or video first. In deep learning, convolutional neural networks are usually used for convolution operations, processing feature semantic information pixel by pixel, and finding similarities with labels. naturalness. Therefore, the semantic composition of the image and the correlation between each pixel determine the efficiency and complexity of the convolution operation. [0003] Do...

AI Technical Summary

Problems solved by technology

[0004] At present, some scholars have proposed shadow detection algorithms, but most of them do not take into account the applic

Images