[ICCV 2023] Guiding Local Feature Matching with Surface Curvature

Shuzhe Wang, Juho Kannala, Marc Pollefeys, Daniel Barath

We propose a new method, called curvature similarity extractor (CSE), for improving local feature matching across images. CSE calculates the curvature of the local 3D surface patch for each detected feature point in a viewpoint-invariant manner via fitting quadrics to predicted monocular depth maps. This curvature is then leveraged as an additional signal in feature matching with off-the-shelf matchers like SuperGlue and LoFTR. Additionally, CSE enables end-to-end joint training by connecting the matcher and depth predictor networks. Our experiments demonstrate on large-scale real-world datasets that CSE consistently improves the accuracy of state-of-the-art methods. Fine-tuning the depth prediction network further enhances the accuracy. The proposed approach achieves state-of-the-art results on the ScanNet dataset, showcasing the effectiveness of incorporating 3D geometric information into feature matching.

Environment

Our curvature similarity extractor is an add-on component for advanced matchers. Here we consider the QuadTree as the matcher and DPT for the depth estimation. Please consider setup the QuadTree environment with the following commands:

git clone [email protected]:Tangshitao/QuadTreeAttention.git
cd QuadTreeAttention&&python setup.py install

Download our CSE module and setup the environment with the following commands.

cd ..
git clone [email protected]:AaltoVision/surface-curvature-estimator.git
cd surface-curvature-estimator
conda env create -f environment.yaml
conda activate 

Dataset

For Megadepth and ScanNet datasets, please refer to the LoFTR for dataset setup. For YFCC100M, you can use the OANet to download it.

cd ..
git clone https:/zjhthu/OANet
cd OANet
bash download_data.sh raw_data raw_data_yfcc.tar.gz 0 8
tar -xvf raw_data_yfcc.tar.gz
# YFCC100M
ln -s raw_data/yfcc100m/* /path_to/data/yfcc/test

Evaluation

We provide the evaluation without any model fine-tuning. The batch size is set to 1 to allow single gpu evaluation. Please consider downloading the models here and following the commands below for the evaluation. The weights can also be downloaded from the original QuadTree and DPT repos. Note we run the evaluation with and without CSE for a more fair comparison.

# For ScanNet
sh scripts/reproduce_test/indoor_ds_quadtree_with_cse.sh
# For MegaDepth
sh scripts/reproduce_test/outdoor_ds_quadtree_with_cse_MEGA.sh
# For YFCC
scripts/reproduce_test/outdoor_ds_quadtree_with_cse_YFCC.sh

Acknowledgements

We appreciate the previous open-source repositories QuadTree , LoFTR , and DPT

Citation

Please consider citing our papers if you find this code useful for your research:

@InProceedings{Wang_2023_ICCV,
    author    = {Wang, Shuzhe and Kannala, Juho and Pollefeys, Marc and Barath, Daniel},
    title     = {Guiding Local Feature Matching with Surface Curvature},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2023},
    pages     = {17981-17991}
}