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

As the scale of object detection dataset is smaller than that of imagerecognition dataset ImageNet, transfer learning has become a basic trainingmethod for deep learning object detection models, which will pretrain thebackbone network of object detection model on ImageNet dataset to extractfeatures for classification and localization subtasks. However, theclassification task focuses on the salient region features of object, while thelocation task focuses on the edge features of object, so there is certaindeviation between the features extracted by pretrained backbone network and thefeatures used for localization task. In order to solve this problem, adecoupled self attention(DSA) module is proposed for one stage object detectionmodels in this paper. DSA includes two decoupled self-attention branches, so itcan extract appropriate features for different tasks. It is located between FPNand head networks of subtasks, so it is used to extract global features basedon FPN fused features for different tasks independently. Although the networkof DSA module is simple, but it can effectively improve the performance ofobject detection, also it can be easily embedded in many detection models. Ourexperiments are based on the representative one-stage detection modelRetinaNet. In COCO dataset, when ResNet50 and ResNet101 are used as backbonenetworks, the detection performances can be increased by 0.4% AP and 0.5% APrespectively. When DSA module and object confidence task are applied inRetinaNet together, the detection performances based on ResNet50 and ResNet101can be increased by 1.0% AP and 1.4% AP respectively. The experiment resultsshow the effectiveness of DSA module. Code is at:https://github.com/chenzuge1/DSANet.git.

Decoupled Self Attention for Accurate One Stage Object Detection