English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

In this paper, we focus on exploring effective methods for faster andaccurate semantic segmentation. A common practice to improve the performance isto attain high-resolution feature maps with strong semantic representation. Twostrategies are widely used: atrous convolutions and feature pyramid fusion,while both are either computationally intensive or ineffective. Inspired by theOptical Flow for motion alignment between adjacent video frames, we propose aFlow Alignment Module (FAM) to learn \textit{Semantic Flow} between featuremaps of adjacent levels and broadcast high-level features to high-resolutionfeatures effectively and efficiently. Furthermore, integrating our FAM to astandard feature pyramid structure exhibits superior performance over otherreal-time methods, even on lightweight backbone networks, such as ResNet-18 andDFNet. Then to further speed up the inference procedure, we also present anovel Gated Dual Flow Alignment Module to directly align high-resolutionfeature maps and low-resolution feature maps where we term the improved versionnetwork as SFNet-Lite. Extensive experiments are conducted on severalchallenging datasets, where results show the effectiveness of both SFNet andSFNet-Lite. In particular, when using Cityscapes test set, the SFNet-Liteseries achieve 80.1 mIoU while running at 60 FPS using ResNet-18 backbone and78.8 mIoU while running at 120 FPS using STDC backbone on RTX-3090. Moreover,we unify four challenging driving datasets into one large dataset, which wenamed Unified Driving Segmentation (UDS) dataset. It contains diverse domainand style information. We benchmark several representative works on UDS. BothSFNet and SFNet-Lite still achieve the best speed and accuracy trade-off onUDS, which serves as a strong baseline in such a challenging setting. The codeand models are publicly available at https://github.com/lxtGH/SFSegNets.

SFNet: Faster and Accurate Semantic Segmentation via Semantic Flow