This paper introduces SuperPoint, a self-supervised framework for training interest point detectors and descriptors suitable for multiple-view geometry problems in computer vision. Unlike patch-based neural networks, SuperPoint operates on full-sized images and jointly computes pixel-level interest point locations and associated descriptors in a single forward pass. The authors introduce Homographic Adaptation, a multi-scale, multi-homography approach to enhance interest point detection repeatability and perform cross-domain adaptation. When trained on the MS-COCO dataset using Homographic Adaptation, SuperPoint outperforms traditional detectors and other methods in terms of repeatability and homography estimation on the HPatches dataset. The final system demonstrates state-of-the-art performance in homography estimation compared to LIFT, SIFT, and ORB.This paper introduces SuperPoint, a self-supervised framework for training interest point detectors and descriptors suitable for multiple-view geometry problems in computer vision. Unlike patch-based neural networks, SuperPoint operates on full-sized images and jointly computes pixel-level interest point locations and associated descriptors in a single forward pass. The authors introduce Homographic Adaptation, a multi-scale, multi-homography approach to enhance interest point detection repeatability and perform cross-domain adaptation. When trained on the MS-COCO dataset using Homographic Adaptation, SuperPoint outperforms traditional detectors and other methods in terms of repeatability and homography estimation on the HPatches dataset. The final system demonstrates state-of-the-art performance in homography estimation compared to LIFT, SIFT, and ORB.