Faster and more accurate scene segmentation by being aware of the decoder stages with SFANet
Faster and more accurate scene segmentation by being aware of the decoder stages with SFANet
Stage-Aware Feature Alignment Network for Real-Time Semantic Segmentation of Street Scenes
arXiv paper abstract https://arxiv.org/abs/2203.04031v1
arXiv PDF paper https://arxiv.org/pdf/2203.04031v1.pdf
Over the past few years, deep convolutional neural network-based methods have made great progress in semantic segmentation of street scenes.
Some recent methods align feature maps … ignore the different roles of stages … greatly affects the inference speed.
… present a novel Stage-aware Feature Alignment Network (SFANet) based on the encoder-decoder structure for real-time semantic segmentation of street scenes.
… Stage-aware Feature Alignment module (SFA) is proposed to align and aggregate two adjacent levels of feature maps effectively.
… by taking into account the unique role of each stage in the decoder, a novel stage-aware Feature Enhancement Block (FEB) is designed to enhance spatial details and contextual information of feature maps from the encoder.
… SFANet respectively obtains 78.1% and 74.7% … (mIoU) at … 37 FPS and 96 FPS on the … Cityscapes and CamVid test datasets … using … a single GTX 1080Ti GPU.
Stay up to date. Subscribe to my posts https://morrislee1234.wixsite.com/website/contact
Web site with my other posts by category https://morrislee1234.wixsite.com/website
LinkedIn https://www.linkedin.com/in/morris-lee-47877b7b
