search

liutinglt / CE2P

214 stars 41 forks source link

cannot reproduce the results as you mentioned #7

Open lyxlynn opened 5 years ago

lyxlynn commented 5 years ago

Hi I use your code 'job_train.sh' to train the LIP images and the same configs as you. when I test the testimages, the accuracies are: 78.35 | 43.73 | 35.83 | 64.27 the mean IOU on valimages: 0.41 The results are much lower than yours.

eng100200 commented 5 years ago

@Liuyixuan95 can i get your email ID?

lyxlynn commented 5 years ago

@eng100200 sorry I just see it liuyixua17@mails.tsinghua.edu.cn

mxtx0509 commented 5 years ago

Hi, I run job_evaluate.sh code using LIP_CE2P_train.pth model on LIP val set and obtain very poor results: Pixel accuracy: 0.769786 Mean accuracy: 0.451063 Mean IU: 0.308818 Why are the results much lower than yours? Our settings: GPU: V100, Cuda: 9.0, pytorch: 0.3

yellowoo commented 5 years ago

I have exactly the same result as @mxtx0509. My settings: GPU 1080Ti, Cuda 9.0, pytorch 0.3.1, python 3.5/3.6, opencv 4.0.0

YU-0626 commented 3 years ago

I have exactly the same result as @mxtx0509. My settings: GPU 1080Ti, Cuda 9.0, pytorch 0.3.1, python 3.5/3.6, opencv 4.0.0

Can you tell me your parameter Settings,thanks。