Extreme Relative Pose Network under Hybrid Representations
This improves multi-scan reconstruction in few-view settings for computer vision applications, but is incremental as it builds on existing scene completion and matching techniques.
The paper tackles relative pose estimation for small-overlapping or non-overlapping RGB-D scans by using hybrid representations (360-image, 2D layout, planar patches) and a global-to-local matching procedure, achieving rotation errors of 28.6°/16.8° and translation errors of 0.90m/0.76m on ScanNet for top-1/top-5 predictions.
In this paper, we introduce a novel RGB-D based relative pose estimation approach that is suitable for small-overlapping or non-overlapping scans and can output multiple relative poses. Our method performs scene completion and matches the completed scans. However, instead of using a fixed representation for completion, the key idea is to utilize hybrid representations that combine 360-image, 2D image-based layout, and planar patches. This approach offers adaptively feature representations for relative pose estimation. Besides, we introduce a global-2-local matching procedure, which utilizes initial relative poses obtained during the global phase to detect and then integrate geometric relations for pose refinement. Experimental results justify the potential of this approach across a wide range of benchmark datasets. For example, on ScanNet, the rotation translation errors of the top-1/top-5 predictions of our approach are 28.6/0.90m and 16.8/0.76m, respectively. Our approach also considerably boosts the performance of multi-scan reconstruction in few-view reconstruction settings.