iSAID dataset IoU/ACC nan

[lypan1208]

I tried pspnet on iSAID dataset, as the instruction, I ran tools/dataset_converters/isaid.py to get the dataset as follows:

├── data
│ ├── iSAID
│ │ ├── img_dir
│ │ │ ├── train
│ │ │ ├── val
│ │ │ ├── test
│ │ ├── ann_dir
│ │ │ ├── train
│ │ │ ├── val

then I use the command
python tools/train.py configs/pspnet/pspnet_r50-d8_4xb4-80k_isaid-896x896.py

the model started to train, but the validation got something wrong:
2024/08/22 10:32:06 - mmengine - INFO - Iter(val) [11600/11644] eta: 0:00:02 time: 0.0648 data_time: 0.0024 memory: 1224
2024/08/22 10:32:09 - mmengine - INFO - per class results:
2024/08/22 10:32:09 - mmengine - INFO -
+--------------------+-------+-------+
| Class | IoU | Acc |
+--------------------+-------+-------+
| background | 96.65 | 100.0 |
| ship | 0.0 | 0.0 |
| store_tank | nan | nan |
| baseball_diamond | nan | nan |
| tennis_court | nan | nan |
| basketball_court | nan | nan |
| Ground_Track_Field | nan | nan |
| Bridge | nan | nan |
| Large_Vehicle | nan | nan |
| Small_Vehicle | nan | nan |
| Helicopter | nan | nan |
| Swimming_pool | nan | nan |
| Roundabout | nan | nan |
| Soccer_ball_field | nan | nan |
| plane | nan | nan |
| Harbor | nan | nan |
+--------------------+-------+-------+

mmsegmentation/mmseg/datasets/isaid.py
`import mmengine.fileio as fileio

from mmseg.registry import DATASETS
from .basesegdataset import BaseSegDataset

@DATASETS.register_module()
class iSAIDDataset(BaseSegDataset):
""" iSAID: A Large-scale Dataset for Instance Segmentation in Aerial Images
In segmentation map annotation for iSAID dataset, which is included
in 16 categories. reduce_zero_label is fixed to False. The
img_suffix is fixed to '.png' and seg_map_suffix is fixed to
'_manual1.png'.
"""

METAINFO = dict(
    classes=('background', 'ship', 'store_tank', 'baseball_diamond',
             'tennis_court', 'basketball_court', 'Ground_Track_Field',
             'Bridge', 'Large_Vehicle', 'Small_Vehicle', 'Helicopter',
             'Swimming_pool', 'Roundabout', 'Soccer_ball_field', 'plane',
             'Harbor'),
    palette=[[0, 0, 0], [0, 0, 63], [0, 63, 63], [0, 63, 0], [0, 63, 127],
             [0, 63, 191], [0, 63, 255], [0, 127, 63], [0, 127, 127],
             [0, 0, 127], [0, 0, 191], [0, 0, 255], [0, 191, 127],
             [0, 127, 191], [0, 127, 255], [0, 100, 155]])

def __init__(self,
             img_suffix='.png',
             seg_map_suffix='_instance_color_RGB.png',
             reduce_zero_lavel=True,
             ignore_index=255,
             **kwargs) -> None:
    super().__init__(
        img_suffix=img_suffix,
        seg_map_suffix=seg_map_suffix,
        ignore_index=ignore_index,
        **kwargs)
    assert fileio.exists(
        self.data_prefix['img_path'], backend_args=self.backend_args)

`