Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

The training parameters of S3DIS #14

Open
M-leng opened this issue Oct 25, 2021 · 3 comments
Open

The training parameters of S3DIS #14

M-leng opened this issue Oct 25, 2021 · 3 comments

Comments

@M-leng
Copy link

M-leng commented Oct 25, 2021

Hello, I have followed your tips, and trained the model. On Area 5, S3DIS dataset, the model I trained only achieved 61.078 mIOU. The mIoU in your paper is 65.4. Could your share your training parameters on S3DIS dataset? Looking forward to your reply.

@ShiQiu0419
Copy link
Owner

ShiQiu0419 commented Oct 25, 2021

Hi, is 61.078 your validation result (a sampled data of Area 5) during training? i.e, have you done the post-test on whole Area 5 by running something like:
python -B main_S3DIS.py --gpu 0 --mode test --test_area 5 --model_path 'pretrained/Area5/snap-32251'

I remember that the validation result of my reported model is about 62-63, while after running the above testing command, the final results would be around 65.

The original training log is no longer available on our server since it has been periodically cleaned-up. If I did not made any "silly" mistake, I leave all setting on github as it was on our server. However, it is possible that deep learning always produces some randomness, would you mind running it for multiple times? Alternatively, you can try to modify the batch size and train_steps in helper_tool.py

@raoumairwaheed
Copy link

Hello, I have trained the network with the settings you provided in your code. and after running the test command I got the following results on the areas below. Can you please share your training and testing settings?

Area | Results (me) |Your Results

area2 | 52.98 | 57.8
area3 | 78.83 | 80.0
area4 | 59.77 | 64.3
area5 | 63.51 | 65.4
area6 | 79.98 | 81.8

@ShiQiu0419
Copy link
Owner

ShiQiu0419 commented Feb 24, 2022 via email

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants