感谢你分享的代码，我成功的对一些论文进行了复现，我有几个问题想请教一下您
一、我在LEVIR上复现的18年FC_EF的代码，发现40000次效果好的特别多，下面是结果，这是数据集造成的吗？
+-----------+--------+-----------+--------+-------+-------+
| Class | Fscore | Precision | Recall | IoU | Acc |
+-----------+--------+-----------+--------+-------+-------+
| unchanged | 98.98 | 99.22 | 98.75 | 97.99 | 98.75 |
| changed | 78.04 | 74.22 | 82.27 | 63.98 | 82.27 |
+-----------+--------+-----------+--------+-------+-------+
08/13 10:26:46 - mmengine - INFO - Iter(val) [64/64] aAcc: 98.0600 mFscore: 88.5100 mPrecision: 86.7200 mRecall: 90.5100 mIoU: 80.9900 mAcc: 90.5100 data_time: 0.0056 time: 0.0434
08/13 10:26:46 - mmengine - INFO - The previous best checkpoint /hy-tmp/open-cd/fc_ef_256x256_40k_levircd_workdir/best_mIoU_iter_36000.pth is removed
08/13 10:26:47 - mmengine - INFO - The best checkpoint with 80.9900 mIoU at 40000 iter is saved to best_mIoU_iter_40000.pth.
二、我打算用其他数据集像WHU/SYSU这种，我已经把它们处理成256*256的了，我只在LEVIR的配置上修改下数据集路径可以去用这些数据集去复现吗，还是我还要修改下size吗？如果要修改，我在哪修改呢？