Poor F1 score and precision

[wawatsmart]

I run #!/usr/bin/env bash config_path='st.lovecs.2CZ.lovecs' python LoveCS_train.py --config_path=${config_path}

But after that, i got poor f1 score. precision, and recall. What should I fix. Is there anything that I should setting? Why I do not have f1 score. precision, and recall like in the paper?

For replace the batch normalizations with cross-sensor normalizations, I thing it has been on LoveCS_train.py in your github, so I do not modification anything.

[Junjue-Wang]

Yes, and what dataset do you use?

[wawatsmart]

I wanna ask again about dataset. In loveda repositoty there is rural and urban data. But in yours config.py it tell that there is source_dir and target_dir.

source_dir = dict( image_dir=[ './cross_sensor/src_domain/images/', ], mask_dir=[ './cross_sensor/src_domain/masks/', ], ) target_dir = dict( image_dir=[ './cross_sensor/tgt_domain/images/', ], mask_dir=[ './cross_sensor/tgt_domain/masks/', ], )

Is rural and urban be combine or what? Can you tell me which one in loveda become source of the source_dir is and target_dir in lovecs. I suggest I get the poor f1score because wrong configuration in dataset.

[Junjue-Wang]

The LoveDA dataset is not the same as in LoveCS. The first aims to advance domain adaptation between urban and rual and the last is designed between the airborne and spaceborne.

[wawatsmart]

How i can download the same dataset that use in lovecs that has advance domain adaptation . I don't know the way to make advance domain adaptation and designed become airbone and spaceborne from loveda.

[Junjue-Wang]

Sorry, the cross-sensor dataset is not available now. You can use the LoveDA dataset to develop your algorithms first.

[Junjue-Wang]

If you use the LoveDA to develop DA methods. You can try to modify my configs in LoveCS method, i.e, increase the training steps, batch size. Or you can refer to the configs w.r.t this

[spwidodo]

Hi.. Junjue-Wang

if I will use loveDA as a dataset. Is the configuration correct like this:

source_dir = dict( image_dir=[ './LoveDA/Train/Rural/images_png/', ], mask_dir=[ './LoveDA/Train/Rural/masks_png/', ], ) target_dir = dict( image_dir=[ './LoveDA/Val/Rural/images_png/', ], mask_dir=[ './LoveDA/Val/Rural/masks_png/', ], )

or

source_dir = dict( image_dir=[ './LoveDA/Train/Urban/images_png/', ], mask_dir=[ './LoveDA/Train/Urban/masks_png/', ], ) target_dir = dict( image_dir=[ './LoveDA/Val/Rural/images_png/', ], mask_dir=[ './LoveDA/Val/Rural/masks_png/', ], )

thanks..

[Junjue-Wang]

The second setting is right.

Hi.. Junjue-Wang

if I will use loveDA as a dataset. Is the configuration correct like this:

source_dir = dict( image_dir=[ './LoveDA/Train/Rural/images_png/', ], mask_dir=[ './LoveDA/Train/Rural/masks_png/', ], ) target_dir = dict( image_dir=[ './LoveDA/Val/Rural/images_png/', ], mask_dir=[ './LoveDA/Val/Rural/masks_png/', ], )

or

source_dir = dict( image_dir=[ './LoveDA/Train/Urban/images_png/', ], mask_dir=[ './LoveDA/Train/Urban/masks_png/', ], ) target_dir = dict( image_dir=[ './LoveDA/Val/Rural/images_png/', ], mask_dir=[ './LoveDA/Val/Rural/masks_png/', ], )

thanks..

[spwidodo]

The second setting is right.

Hi.. Junjue-Wang if I will use loveDA as a dataset. Is the configuration correct like this: source_dir = dict( image_dir=[ './LoveDA/Train/Rural/images_png/', ], mask_dir=[ './LoveDA/Train/Rural/masks_png/', ], ) target_dir = dict( image_dir=[ './LoveDA/Val/Rural/images_png/', ], mask_dir=[ './LoveDA/Val/Rural/masks_png/', ], ) or source_dir = dict( image_dir=[ './LoveDA/Train/Urban/images_png/', ], mask_dir=[ './LoveDA/Train/Urban/masks_png/', ], ) target_dir = dict( image_dir=[ './LoveDA/Val/Rural/images_png/', ], mask_dir=[ './LoveDA/Val/Rural/masks_png/', ], ) thanks..

thanks for your response.

[rongtongxueya]

May I ask where the data set of the experiment was downloaded? I want to see the situation of this code, but I did not see the introduction of the data set

[wawatsmart]

The problem has been solved. Thank you very much.

On Sat, 2 Sep 2023, 18:47 rongtongxueya, @.***> wrote:

May I ask where the data set of the experiment was downloaded? I want to see the situation of this code, but I did not see the introduction of the data set

— Reply to this email directly, view it on GitHub https://github.com/Junjue-Wang/LoveCS/issues/3#issuecomment-1703811264, or unsubscribe https://github.com/notifications/unsubscribe-auth/AYNJJKMKBQBK5BQOUYJZP6LXYMMDBANCNFSM6AAAAAAQ6COOVE . You are receiving this because you authored the thread.Message ID: @.***>