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

Identifying the camera pose for a given image is a challenging problem withapplications in robotics, autonomous vehicles, and augmented/virtual reality.Lately, learning-based methods have shown to be effective for absolute camerapose estimation. However, these methods are not accurate when generalizing todifferent domains. In this paper, a domain adaptive training framework forabsolute pose regression is introduced. In the proposed framework, the sceneimage is augmented for different domains by using generative methods to trainparallel branches using Barlow Twins objective. The parallel branches leveragea lightweight CNN-based absolute pose regressor architecture. Further, theefficacy of incorporating spatial and channel-wise attention in the regressionhead for rotation prediction is investigated. Our method is evaluated with twodatasets, Cambridge landmarks and 7Scenes. The results demonstrate that, evenwith using roughly 24 times fewer FLOPs, 12 times fewer activations, and 5times fewer parameters than MS-Transformer, our approach outperforms all theCNN-based architectures and achieves performance comparable totransformer-based architectures. Our method ranks 2nd and 4th with theCambridge Landmarks and 7Scenes datasets, respectively. In addition, foraugmented domains not encountered during training, our approach significantlyoutperforms the MS-transformer. Furthermore, it is shown that our domainadaptive framework achieves better performance than the single branch modeltrained with the identical CNN backbone with all instances of the unseendistribution.

A Lightweight Domain Adaptive Absolute Pose Regressor Using Barlow Twins Objective