How to SOLVE this Problem? Remove weird GT visualization (Background and the white rectangle)?

[bjzhb666]

The left is GT, the Visualization of GT is weird.

dataset_type = 'SatelliteSegDataset'
# data_root = 'data/satellite/'
data_root = 'data/satellite_seg_512/'
crop_size = (2048, 2048) # random crop size
bezier_order = 3
dilate_kernel = 1

train_pipeline = [
    dict(type='LoadImageFromFile'), # 第1个流程，从文件路径里加载图像
    dict(type='LoadAnnotations'),  # 第2个流程，对于当前图像，加载它的标注图像
    
    dict(type='RandomResize',  # 调整输入图像大小(resize)和其标注图像的数据增广流程
    scale=crop_size,
    keep_ratio=True),  # 图像裁剪的大小
  
    dict(type='RandomFlip',  # 翻转图像和其标注图像的数据增广流程
        prob=0.5),  # 翻转图像的概率
   
    dict(type='PhotoMetricDistortion'),  # 光学上使用一些方法扭曲当前图像和其标注图像的数据增广流程，和GT无关，不用改
    dict(type='PackSegInputs')  # 打包用于语义分割的输入数据
]

test_pipeline = [
    dict(type='LoadImageFromFile'),  # 第1个流程，从文件路径里加载图像
   
    dict(type='LoadAnnotations'),  # 加载数据集提供的语义分割标注
   
    dict(type='PackSegInputs')  # 打包用于语义分割的输入数据
]

[bjzhb666]

#3649

[Saillxl]

How did you solve it, thank you.

[bjzhb666]

How did you solve it, thank you.

Please refer to the PR I mentioned above. #3649