LoveDA icon indicating copy to clipboard operation
LoveDA copied to clipboard

[NeurIPS 2021] LoveDA: A Remote Sensing Land-Cover Dataset for Domain Adaptive Semantic Segmentation

Results 39 LoveDA issues
Sort by recently updated
recently updated
newest added

Hello, @Junjue-Wang. I want to know whether you have rewrited the dict data type. I can't accessed the members of dict by dot '.'. For example, for the SOURCE_DATA_CONFIG, when...

question

Hello! Wang, the 'config_path' is needed ,how to solve it

question

When I execute bash ./scripts/predict_cbst.sh ,The following error occurred: Traceback (most recent call last): File "predict.py", line 1, in from data.loveda import LoveDALoader File "/home/chen/experiment/remote_sensing/Unsupervised_Domian_Adaptation/data/loveda.py", line 9, in from ever.interface...

question

Hi! I was wondering whether you can advise me on using the model on datasets with input images having more than 3 channels and with image size different from 1024-by-1024....

question

Hi, The mean IoU of our UDA method was 0.446 on the Urban validation data. But we got a score 0 on the Urban test data after submitting test results...

question

Hello, @Junjue-Wang , I notice that in the config file, the image_dir is the same for target_data_config and eval_data_config. I think the image_dir for target_data_config should be the training set...

question

![image](https://user-images.githubusercontent.com/19429672/147403467-f292bd31-63da-4f19-b663-a51757eb3c22.png) When will a result come out

question

thank for your great works. but why the mask is all black pic, i use the weights you provide to test ,output is also a black pic. could u tell...

question

Hi, I have trouble in reproducing the IAST for urban to rural. I found that the data in the rural val dataset are very strange, which has a great gap...

question

Hi, I was wondering what is the difference between _no-data_ and _background_ classes in terms of their meaning (mentioned here: https://github.com/Junjue-Wang/LoveDA#dataset-and-contest), assuming that _background_ means land-cover classes that are not...

question