A Low-light Image Semantic Segmentation Method Based on VSA-DANet

     Semantic segmentation, as one of the fundamental tasks in computer vision, plays an important role in many practical applications. Due to insufficient illumination in low-light images, phenomena such as low brightness, contrast, and signal-to-noise ratio occur, making semantic segmentation of low-light images more difficult. To address this issue, this paper proposes a low-light image semantic segmentation method based on VSA-DANet(Vector Symbolic Architectures-Domain Adaptive Network). VSA-DANet first integrates the image translation network based on vector symbolic architectures and the domain adaptive semantic segmentation network.The image translation network translates normal illumination images into low-light images, reducing style differences between domains and improving the segmentation accuracy of low-light images. Then, a Layered Feature Mapping(LFM)module is proposed in the image translation network. The LFM module can better allocate the super-vectors of vector symbolic architectures to the feature vectors of different layers in the semantic segmentation network, making the translated images more similar to low-light images, thus improving the segmentation accuracy of the segmentation network. Finally, a Cross Domain Rare Class Mixing(CDRCM)method is proposed in the domain adaptive semantic segmentation network. CDRCM obtains rare classes in the low-light domain based on the pseudo-label distribution of low-light images during domain adaptation, and biases towards these rare classes during cross-domain mixing, thereby improving the segmentation accuracy of rare classes in the low-light domain. Experimental results on Cityscapes→Dark Zurich and Cityscapes→ACDC-night show that our method has better improvements in terms of mean Intersection over Union compared to the current mainstream low-light image semantic segmentation methods.