We introduce a novel CNN-based feature point detector - GLAMpoints - learned in a semi-supervised manner. Our detector extracts repeatable, stable interest points with a dense coverage, specifically designed to maximize the correct matching in a specific domain, which is in contrast to conventional techniques that optimize indirect metrics. In this paper, we apply our method on challenging retinal slitlamp images, for which classical detectors yield unsatisfactory results due to low image quality and insufficient amount of low-level features. We show that GLAMpoints significantly outperforms classical detectors as well as state-of-the-art CNN-based methods in matching and registration quality for retinal images. Our method can also be extended to other domains, such as natural images. Training code and model weights are available at https://github.com/PruneTruong/GLAMpoints_pytorch.
updated: Mon Jun 15 2020 17:34:03 GMT+0000 (UTC)
published: Mon Aug 19 2019 14:06:27 GMT+0000 (UTC)