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/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

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

Zendy Tools

Zendy Open

Semantic Segmentation (SS) is the task to assign a semantic label to eachpixel of the observed images, which is of crucial significance for autonomousvehicles, navigation assistance systems for the visually impaired, andaugmented reality devices. However, there is still a long way for SS to be putinto practice as there are two essential challenges that need to be addressed:efficiency and evaluation criterions for practical application. For specificapplication scenarios, different criterions need to be adopted. Recall rate isan important criterion for many tasks like autonomous vehicles. For autonomousvehicles, we need to focus on the detection of the traffic objects like cars,buses, and pedestrians, which should be detected with high recall rates. Inother words, it is preferable to detect it wrongly than miss it, because theother traffic objects will be dangerous if the algorithm miss them and segmentthem as safe roadways. In this paper, our main goal is to explore possiblemethods to attain high recall rate. Firstly, we propose a real-time SS networknamed Swift Factorized Network (SFN). The proposed network is adapted fromSwiftNet, whose structure is a typical U-shape structure with lateralconnections. Inspired by ERFNet and Global convolution Networks (GCNet), wepropose two different blocks to enlarge valid receptive field. They do not takeup too much calculation resources, but significantly enhance the performancecompared with the baseline network. Secondly, we explore three ways to achievehigher recall rate, i.e. loss function, classifier and decision rules. Weperform a comprehensive set of experiments on state-of-the-art datasetsincluding CamVid and Cityscapes. We demonstrate that our SS convolutionalneural networks reach excellent performance. Furthermore, we make a detailedanalysis and comparison of the three proposed methods on the promotion ofrecall rate.

A Comparative Study of High-Recall Real-Time Semantic Segmentation Based on Swift Factorized Network