This is the official implement of the RWD-Net from the paper Measuring the rogue wave pattern triggered from Gaussian perturbations by deep learning.

Python 3.7

Tensorflow-gpu==1.13.1

Keras==2.1.5

We release the RWD-10K dataset which has 10191 rogue wave images. All the images are named as aXeYuZ where X, Y and Z are the orresponding parameter values in the initial equation. One .jpg image file corresponds to one .xml file which contains the bounding boxes annotation of the origin images for the rogue wave detection. You can see more details about the dataset in our paper above. You can download the RWD-10K dataset here. If you use this dataset for your research, please cite our paper.

Once you download the RWD-10K dataset, create the following folders and put the images at RogueWaves/images and put the xml files at RogueWaves/xmls and run

python 0_gencsv.py

And the data splits are saved in RogueWave/Annotations.

cd RogueWave/ and run

python 1_train.py

And the trained models are saved in RogueWaves/snapshots.

once you cd RogueWave/ and run

python 2_convert.py

to convert the model saved in RogueWaves/model for testing.

run

python 3_test2input.py

for putting the test images in RogueWave/input, then run

python 4_test.py

for testing. Finally, for evalutation you can run

python 5_eval.py

For inference, you can run

python 6_predict.py

for detecting your images.

@article{zou2021measuring,
  title={Measuring the rogue wave pattern triggered from Gaussian perturbations by deep learning},
  author={Liwen Zou, XinHang Luo, Delu Zeng, Liming Ling and Li-Chen Zhao},
  journal={arXiv preprint arXiv:2109.08909},
  year={2021}
}

Part of codes are reused from the SKU110K. Thanks to Eran Goldman et al. for the codes of SKU-110K detector.

Liwen Zou([email protected])