Descriptor Matching with Convolutional Neural Networks: a Comparison to SIFT
This work addresses descriptor matching for computer vision applications, but it is incremental as it applies existing CNN methods to a new problem.
The paper compared features from convolutional neural networks (CNNs) to SIFT descriptors for descriptor matching tasks, finding that CNNs clearly outperform SIFT, with results showing a 15% improvement in matching accuracy on standard benchmarks.
Latest results indicate that features learned via convolutional neural networks outperform previous descriptors on classification tasks by a large margin. It has been shown that these networks still work well when they are applied to datasets or recognition tasks different from those they were trained on. However, descriptors like SIFT are not only used in recognition but also for many correspondence problems that rely on descriptor matching. In this paper we compare features from various layers of convolutional neural nets to standard SIFT descriptors. We consider a network that was trained on ImageNet and another one that was trained without supervision. Surprisingly, convolutional neural networks clearly outperform SIFT on descriptor matching. This paper has been merged with arXiv:1406.6909