Code for "LoFTR: Detector-Free Local Feature Matching with Transformers", CVPR 2021, T-PAMI 2022

🤗 image matching webui

Code for "Efficient LoFTR: Semi-Dense Local Feature Matching with Sparse-Like Speed", CVPR 2024

Multiview matching with deep-learning and hand-crafted local features for COLMAP and other SfM software. Supports high-resolution formats and images with rotations. Both CLI and GUI are supported.

small c++ library to quickly deploy models using onnxruntime

Distilled coarse part of LoFTR adapted for compatibility with TensorRT and embedded divices

一个利用深度学习进行特征提取匹配的图像拼接应用

ROB 530 Final Project

Pose estimation pipeline for 3D Reconstruction using LoFTR (Local Feature Transformer) detector free feature matcher.

No description available

CVPR 2022 "Image Matching: Local Features and Beyond" workshop challenge: Silver Medal 🥈

LoFTR:Detector-Free Local Feature Matching with Transformers CVPR 2021(with train script --- unofficial ---)

No description available

Translate loftr indoor and outdoor weights to onnx model

[CVPR'22 Image Matching Workshop (IMW) Top6%] Propose a model registering two images from diverse viewpoints with coarse-to-fine attention, use LoFTR for local feature matching and DKM for regression with deep kernels and Gaussian processes.

Deep Electron Microscopy Image Stitching (DEMIS): a tool for stitching grids of electron microscopy images. Image matching is performed using LoFTR.

TensorRT adaptation of LoFTR: Detector-Free Local Feature Matching with Transformers

No description available

Code for our re-implementation of "LoFTR: Detector-Free Local Feature Matching with Transformers"

Reimplementation of LoFTR model and training in PyTorch, end-to-end transformer model for image matching

CNN (LoFTR)-based image-matching implementation in a plotly dash dashboard derived from the google image matching challenge 2022 data. See: https://www.kaggle.com/competitions/image-matching-challenge-2022/overview

A hands-on 12-week Robotics and AI integration mini project | Pytorch -> LoFTR style matching -> CUDA

One-command pipeline that turns ordinary photos into precise 3D reconstructions using LoFTR, SuperGlue, and COLMAP.

Code for "LoFTR: Detector-Free Local Feature Matching with Transformers", CVPR 2021 written in pure Julia

Nix Flake for LoFTR

No description available

Homography-based perspective video transition using LoFTR feature matching and DeepLabV3 person segmentation for seamless foreground-aware stitching between clips.

This work was part of my submission for IMC 22.

Rust workspace for a native tch-based implementation of LoFTR

借用loftr的思想反推pose，作为解决slam问题的新途径