aRI0U / RandLA-Net-pytorch

PyTorch implementation of RandLA-Net

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

about results

LARLSN opened this issue · comments

commented

hello,could you please provide your best result got by your code

Gentle bump on this issue. It would be really helpful to know to which extent the performance are different than the ones in the original paper ! :)

HI @LARLSN ,
on S3DIS dataset i will get this result and it is very differs from original one.

`=== EPOCH 99/100 ===
Training loss: 1.9831392 Validation loss: 680.1278906
Accuracy | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | OA
Training: | 0.927 | 0.091 | 0.316 | 0.962 | 0.671 | 0.000 | 0.000 | 0.041 | 0.781 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.607
Validation: | 0.952 | 0.010 | 0.164 | 0.368 | 0.287 | 0.000 | 0.000 | 0.011 | 0.320 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.304
IoU | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | mIoU
Training: | 0.764 | 0.055 | 0.081 | 0.906 | 0.466 | 0.000 | 0.000 | 0.009 | 0.218 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.322
Validation: | 0.325 | 0.008 | 0.053 | 0.300 | 0.189 | 0.000 | 0.000 | 0.007 | 0.073 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.095
Time elapsed: 3 min 50 s

=== EPOCH 100/100 ===
Training loss: 1.8751107 Validation loss: 17.0391586
Accuracy | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | OA
Training: | 0.917 | 0.041 | 0.312 | 0.961 | 0.637 | 0.000 | 0.000 | 0.276 | 0.664 | 0.000 | 0.000 | 0.000 | 0.001 | 0.000 | 0.580
Validation: | 0.487 | 0.017 | 0.128 | 0.745 | 0.562 | 0.000 | 0.000 | 0.231 | 0.108 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.317
IoU | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | mIoU
Training: | 0.764 | 0.031 | 0.075 | 0.918 | 0.468 | 0.000 | 0.000 | 0.054 | 0.186 | 0.000 | 0.000 | 0.000 | 0.001 | 0.000 | 0.310
Validation: | 0.290 | 0.013 | 0.055 | 0.346 | 0.350 | 0.000 | 0.000 | 0.045 | 0.023 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.119
Time elapsed: 3 min 51 s`

Hello, I and you encountered the same problem, excuse me, how do you finally give up?

@LARLSN,在S3DIS数据集上,我会得到这个结果,它与原始结果非常不同。>> `=== EPOCH 99/100 ===培训损失:1.9831392 验证损失:680.1278906准确性 | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | OA培训: | 0.927 | 0.091 | 0.316 | 0.962 | 0.671 | 0.000 | 0.000 | 0.041 | 0.781 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.607验证: | 0.952 | 0.010 | 0.164 | 0.368 | 0.287 | 0.000 | 0.000 | 0.011 | 0.320 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.304IoU | 0 | 1| 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | mIoU培训: | 0.764 | 0.055 | 0.081 | 0.906 | 0.466 | 0.000 | 0.000 | 0.009 | 0.218 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.322验证: | 0.325 | 0.008 | 0.053 | 0.300 | 0.189 | 0.000 | 0.000 | 0.007 | 0.073 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.095时间:3分50秒>> === EPOCH 100/100 ===培训损失:1.8751107 验证损失:1.0391586准确性 | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | OA培训: | 0.917 | 0.041 | 0.312 | 0.961 | 0.637 | 0.000 | 0.000 | 0.276 | 0.664 | 0.000 | 0.000 | 0.000 | 0.001 | 0.000 | 0.580验证: | 0.487 | 0.017 | 0.128 | 0.745 | 0.562 | 0.000 | 0.000 | 0.231 | 0.108 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.317IoU | 0 | 1| 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | mIoU培训: | 0.764 | 0.031 | 0.075 | 0.918 | 0.468 | 0.000 | 0.000 | 0.054 | 0.186 | 0.000 | 0.000 | 0.000 | 0.001 | 0.000 | 0.310验证: | 0.290 | 0.013 | 0.055 | 0.346 | 0.350 | 0.000 | 0.000 | 0.045 | 0.023 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.119时间:351秒

哈哈,are you solved the problem? I meet the questions, too