回転同変特徴による局所特徴マッチング
ReF -- Rotation Equivariant Features for Local Feature Matching
ステアラブルCNNで回転特異的な特徴を生成し、グループプーリングで回転不変な局所特徴を実現する手法を提案。データ拡張ベースのCNNと組み合わせて最先端の回転ロバストマッチングを達成した。
著者: Abhishek Peri, Kinal Mehta, Avneesh Mishra, Michael Milford, Sourav Garg, K. Madhava Krishna
分類: cs.CV, cs.RO
原文アブストラクト
Sparse local feature matching is pivotal for many computer vision and robotics tasks. To improve their invariance to challenging appearance conditions and viewing angles, and hence their usefulness, existing learning-based methods have primarily focused on data augmentation-based training. In this work, we propose an alternative, complementary approach that centers on inducing bias in the model architecture itself to generate `rotation-specific' features using Steerable E2-CNNs, that are then group-pooled to achieve rotation-invariant local features. We demonstrate that this high performance, rotation-specific coverage from the steerable CNNs can be expanded to all rotation angles by combining it with augmentation-trained standard CNNs which have broader coverage but are often inaccurate, thus creating a state-of-the-art rotation-robust local feature matcher. We benchmark our proposed methods against existing techniques on HPatches and a newly proposed UrbanScenes3D-Air dataset for visual place recognition. Furthermore, we present a detailed analysis of the performance effects of ensembling, robust estimation, network architecture variations, and the use of rotation priors.
