S

Superpoint

Developed by magic-leap-community
SuperPoint is a self-supervised trained fully convolutional network for interest point detection and description.
Downloads 59.12k
Release Time : 3/13/2024

Model Overview

The SuperPoint model can detect repeatable interest points under homography transformations and provides descriptors for each point. It is primarily used as a feature extractor for other tasks such as homography estimation and image matching.

Model Features

Self-supervised Training
The model is trained in a self-supervised manner, eliminating the need for large amounts of labeled data.
Joint Detection and Description
Computes both interest point locations and associated descriptors in a single forward pass.
Homography Adaptation
Uses multi-scale, multi-homography methods to improve the repeatability of interest point detection.

Model Capabilities

Interest Point Detection
Feature Description
Image Matching

Use Cases

Computer Vision
Homography Estimation
Used to estimate homography transformations between images
Achieved state-of-the-art homography estimation results on the HPatches dataset
Image Matching
Performs feature matching between images under different perspectives or conditions
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase