/LSCCL

Primary LanguagePython

Sub-pixel Checkerboard Corner Localization

A novel end-to-end sub-pixel checkerboard corner detection method.

Have fun

  1. Create synthetic dataset:
create_dataset.py

The examples of background and texture images required for synthesizing data are shown in the folder data/creat_dataset. Please add your data.

  1. Train model:

    Due to the limited performance of our GPU, we did not train larger image sizes. If you are interested, you can also use larger image sizes for training.

python trian.py

  1. Test your image:

    We provide training weights so you can test your dataset. It should be noted that if your test image resolution is too large, please zoom in to a width and height of less than 600 pixels first (this will have the best effect).

python demo.py

Citation

If you find this code useful for your research, please use the following BibTeX entry.

H. Zhu, Z. Zhou, B. Liang, X. Han and Y. Tao, "Sub-pixel Checkerboard Corner Localization for Robust Vision Measurement," in IEEE Signal Processing Letters, doi: 10.1109/LSP.2023.3340060.

Our code draws inspiration from 'Learning Multi Instance Sub pixel Point Localization'. Please cite the following paper:

@InProceedings{Schroeter_2020_ACCV,
    author    = {Schroeter, Julien and Tuytelaars, Tinne and Sidorov, Kirill and Marshall, David},
    title     = {Learning Multi-Instance Sub-pixel Point Localization},
    booktitle = {Proceedings of the Asian Conference on Computer Vision (ACCV)},
    month     = {November},
    year      = {2020}