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

Arbitrary-oriented objects widely appear in natural scenes, aerialphotographs, remote sensing images, etc., thus arbitrary-oriented objectdetection has received considerable attention. Many current rotation detectorsuse plenty of anchors with different orientations to achieve spatial alignmentwith ground truth boxes, then Intersection-over-Union (IoU) is applied tosample the positive and negative candidates for training. However, we observethat the selected positive anchors cannot always ensure accurate detectionsafter regression, while some negative samples can achieve accuratelocalization. It indicates that the quality assessment of anchors through IoUis not appropriate, and this further lead to inconsistency betweenclassification confidence and localization accuracy. In this paper, we proposea dynamic anchor learning (DAL) method, which utilizes the newly definedmatching degree to comprehensively evaluate the localization potential of theanchors and carry out a more efficient label assignment process. In this way,the detector can dynamically select high-quality anchors to achieve accurateobject detection, and the divergence between classification and regression willbe alleviated. With the newly introduced DAL, we achieve superior detectionperformance for arbitrary-oriented objects with only a few horizontal presetanchors. Experimental results on three remote sensing datasets HRSC2016, DOTA,UCAS-AOD as well as a scene text dataset ICDAR 2015 show that our methodachieves substantial improvement compared with the baseline model. Besides, ourapproach is also universal for object detection using horizontal bound box. Thecode and models are available at https://github.com/ming71/DAL.

Dynamic Anchor Learning for Arbitrary-Oriented Object Detection