🤖 AI Summary
To address domain shift between training and testing caused by illumination, viewpoint, and environmental variations in natural scenes, this paper proposes the first hypernetwork framework for camera pose regression, enabling input-adaptive dynamic weight generation and compatible with mainstream architectures such as PoseNet and DSAC++. To systematically evaluate cross-appearance robustness, we introduce and publicly release Extended Cambridge Landmarks (ECL), a new large-scale, multi-season dataset exhibiting significant appearance disparities. Extensive experiments on multiple benchmarks demonstrate substantial improvements: average position error is reduced by 12.7%, and orientation error by 9.4%. All code, pre-trained models, and the ECL dataset are fully open-sourced to facilitate reproducible research and community advancement.
📝 Abstract
In this work, we propose HyperPose, which utilizes hyper-networks in absolute camera pose regressors. The inherent appearance variations in natural scenes, attributable to environmental conditions, perspective, and lighting, induce a significant domain disparity between the training and test datasets. This disparity degrades the precision of contemporary localization networks. To mitigate this, we advocate for incorporating hypernetworks into single-scene and multiscene camera pose regression models. During inference, the hypernetwork dynamically computes adaptive weights for the localization regression heads based on the particular input image, effectively narrowing the domain gap. Using indoor and outdoor datasets, we evaluate the HyperPose methodology across multiple established absolute pose regression architectures. We also introduce and share the Extended Cambridge Landmarks (ECL), a novel localization dataset, based on the Cambridge Landmarks dataset, showing it in multiple seasons with significantly varying appearance conditions. Our empirical experiments demonstrate that HyperPose yields notable performance enhancements for single- and multi-scene architectures. We have made our source code, pre-trained models, and the ECL dataset openly available.
