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 Plus annual

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

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

The burgeoning navigation services using digital maps provide greatconvenience to drivers. Nevertheless, the presence of anomalies in lanerendering map images occasionally introduces potential hazards, as suchanomalies can be misleading to human drivers and consequently contribute tounsafe driving conditions. In response to this concern and to accurately andeffectively detect the anomalies, this paper transforms lane rendering imageanomaly detection into a classification problem and proposes a four-phasepipeline consisting of data pre-processing, self-supervised pre-training withthe masked image modeling (MiM) method, customized fine-tuning usingcross-entropy based loss with label smoothing, and post-processing to tackle itleveraging state-of-the-art deep learning techniques, especially thoseinvolving Transformer models. Various experiments verify the effectiveness ofthe proposed pipeline. Results indicate that the proposed pipeline exhibitssuperior performance in lane rendering image anomaly detection, and notably,the self-supervised pre-training with MiM can greatly enhance the detectionaccuracy while significantly reducing the total training time. For instance,employing the Swin Transformer with Uniform Masking as self-supervisedpretraining (Swin-Trans-UM) yielded a heightened accuracy at 94.77% and animproved Area Under The Curve (AUC) score of 0.9743 compared with the pure SwinTransformer without pre-training (Swin-Trans) with an accuracy of 94.01% and anAUC of 0.9498. The fine-tuning epochs were dramatically reduced to 41 from theoriginal 280. In conclusion, the proposed pipeline, with its incorporation ofself-supervised pre-training using MiM and other advanced deep learningtechniques, emerges as a robust solution for enhancing the accuracy andefficiency of lane rendering image anomaly detection in digital navigationsystems.

Intelligent Anomaly Detection for Lane Rendering Using Transformer with Self-Supervised Pre-Training and Customized Fine-Tuning