抱歉打扰了，有两个问题想问一下
1.论文中提到“All data have the unified input size of 224 × 224. We apply simple data augmentations including random rotation and flip.”
关键在于flip,我看提供的代码中只使用了RandomGenerator，没有flip，那最好的成绩83.86使用了flip吗。
2.我用作者提供代码和提供的环境版本，在完全没有改动情况下
Mean class 1 mean_dice : 0.881864 mean_hd95 : 12.379786
Mean class 2 mean_dice : 0.624634 mean_hd95 : 13.372712
Mean class 3 mean_dice : 0.816461 mean_hd95 : 61.673903
Mean class 4 mean_dice : 0.729119 mean_hd95 : 44.397654
Mean class 5 mean_dice : 0.939397 mean_hd95 : 12.745916
Mean class 6 mean_dice : 0.666587 mean_hd95 : 10.225262
Mean class 7 mean_dice : 0.900426 mean_hd95 : 36.049106
Mean class 8 mean_dice : 0.786864 mean_hd95 : 13.125501
Testing performance in best val model: mean_dice : 0.793169 mean_hd95 : 25.496230
跑出来差异较大的成绩，我使用的2张1080ti，150个epoch
抱歉打扰了，有两个问题想问一下
1.论文中提到“All data have the unified input size of 224 × 224. We apply simple data augmentations including random rotation and flip.”
关键在于flip,我看提供的代码中只使用了RandomGenerator，没有flip，那最好的成绩83.86使用了flip吗。
2.我用作者提供代码和提供的环境版本，在完全没有改动情况下
Mean class 1 mean_dice : 0.881864 mean_hd95 : 12.379786
Mean class 2 mean_dice : 0.624634 mean_hd95 : 13.372712
Mean class 3 mean_dice : 0.816461 mean_hd95 : 61.673903
Mean class 4 mean_dice : 0.729119 mean_hd95 : 44.397654
Mean class 5 mean_dice : 0.939397 mean_hd95 : 12.745916
Mean class 6 mean_dice : 0.666587 mean_hd95 : 10.225262
Mean class 7 mean_dice : 0.900426 mean_hd95 : 36.049106
Mean class 8 mean_dice : 0.786864 mean_hd95 : 13.125501
Testing performance in best val model: mean_dice : 0.793169 mean_hd95 : 25.496230
跑出来差异较大的成绩，我使用的2张1080ti，150个epoch