Robust Angular Local Descriptor Learning
This work addresses robustness issues in learned local descriptors for computer vision tasks, representing an incremental improvement over existing methods.
The paper tackled the problems of sensitivity to incorrect correspondences in margin loss and the misuse of L2 distance for normalized descriptors by proposing a robust angular loss, achieving state-of-the-art results on the Brown dataset and excellent generalization on Hpatches and Wide Baseline Stereo datasets.
In recent years, the learned local descriptors have outperformed handcrafted ones by a large margin, due to the powerful deep convolutional neural network architectures such as L2-Net [1] and triplet based metric learning [2]. However, there are two problems in the current methods, which hinders the overall performance. Firstly, the widely-used margin loss is sensitive to incorrect correspondences, which are prevalent in the existing local descriptor learning datasets. Second, the L2 distance ignores the fact that the feature vectors have been normalized to unit norm. To tackle these two problems and further boost the performance, we propose a robust angular loss which 1) uses cosine similarity instead of L2 distance to compare descriptors and 2) relies on a robust loss function that gives smaller penalty to triplets with negative relative similarity. The resulting descriptor shows robustness on different datasets, reaching the state-of-the-art result on Brown dataset , as well as demonstrating excellent generalization ability on the Hpatches dataset and a Wide Baseline Stereo dataset.