Skip to content

drinkingcoder/NeuralMarker

Repository files navigation

NeuralMarker

NeuralMarker: A Framework for Learning General Marker Correspondence
Zhaoyang Huang*, Xiaokun Pan*, Weihong Pan, Weikang Bian, Yan Xu, Ka Chun Cheung, Guofeng Zhang, Hongsheng Li
SIGGRAPH Asia (ToG) 2022

TODO List

  • Code release
  • Models release
  • Demo code release
  • Dataset&Evaluation code release

Environment

conda create -n neuralmarker
conda activate neuralmarker
conda install python=3.7
pip install -r requirements.txt

Dataset

We use the MegaDepth dataset that preprocessed by CAPS, which is provided in this link. We generate FlyingMarkers training set online. To genenerate FlyingMarkers validation set and test set, please execute:

python synthesis_datasets.py --root ./data/MegaDepth_CAPS/ --csv ./data/synthesis_validate_release.csv --save_dir ./data/flyingmarkers/validation
python synthesis_datasets.py --root ./data/MegaDepth_CAPS/ --csv ./data/synthesis_validate_short.csv --save_dir ./data/validation/synthesis
python synthesis_datasets.py --root ./data/MegaDepth_CAPS/ --csv ./data/synthesis_test_release.csv --save_dir ./data/flyingmarkers/test

The pretrained models, DVL-Markers benchmark, and data for demo are stored in Google Drive.

Training

We train our model on 6 V100 with batch size 2.

CUDA_VISIBLE_DEVICES=0,1,2,3,4,5 python train.py

DVL-Markers Evaluation

Put the DVL-Markers dataset in data:

├── data 
    ├── DVL
        ├── D
        ├── V
        ├── L
        ├── marker

then run

bash eval_DVL.sh

The results will be saved in output

FlyingMarkers Evaluation

python evaluation_FM.py

Demo

for video demo, run

bash demo_video.sh

Acknowledgements

We thank Yijin Li, Rensen Xu, and Jundan Luo for their help. We refer DGC-Net to generate synthetic image pairs.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •