feature detection and matching algorithm models

Introduction

🚀🚀This warehouse mainly uses C++ to compare traditional image feature detection and matching, and deep learning feature detection and matching algorithm models. Deep learning includes superpoint-superglue, and traditional algorithms include AKAZE, SURF, ORB, etc.

1. akaze feature point detection and matching display.

2. superpoint-superpoint feature point detection and matching display.

Dependencies

All operating environments, please strictly follow the given configuration，the configuration is as follows：

OpenCV >= 3.4

CUDA >=10.2

CUDNN>=8.02

TensorRT>=7.2.3

How to Run

1. build.

cd feature-detection-matching-algorithm/
mkdir build
cd build
cmake ..
make

2. run camera.

deep learning algorithms.

./IR --deeplearning --camera 0

traditional algorithms.

./IR --traditional  --camera 0

3. run image-pair.

deep learning algorithms.

./IR --deeplearning --image-pair xx01.jpg xx02.jpg

traditional algorithms.

./IR --traditional  --image-pair xx01.jpg xx02.jpg

Models TRT

https://pan.baidu.com/s/1CoK_KuC42BFD-mtO-BBhHg Code：cb7x

TODO

• Optimizing post-processing using custom TensorRT layer or Cublass.
• Model conversion script.
• support for FP16/INT8.

Discussion

WeChat ID: sigma1573

Welcome to add WeChat(note: unit + name) and join the group discussion！

For more details, please refer to zhihu: https://zhuanlan.zhihu.com/p/518877309

SuperPoint

Superpoint pretrained models are from magicleap/SuperPointPretrainedNetwork.

SuperGlue

SuperGlue pretrained models are from magicleap/SuperGluePretrainedNetwork.

Reference

@inproceedings{sarlin20superglue,
  author    = {Paul-Edouard Sarlin and
               Daniel DeTone and
               Tomasz Malisiewicz and
               Andrew Rabinovich},
  title     = {{SuperGlue}: Learning Feature Matching with Graph Neural Networks},
  booktitle = {CVPR},
  year      = {2020},
  url       = {https://arxiv.org/abs/1911.11763}
}