About the pretrained model

[Terminal-K]

Hi, i used the pretrained weights which you provided to predict, and i get the result as follow:

It doesn't seem to be the weight in the paper

[MaxChanger]

Thanks for your interest in our project.
Could you provided more details about the command you used?
And which pretrianed model used for inference ?

[Terminal-K]

The pretrained weights i used is from here

i first inferred on dataset by running

python3 ./infer.py -d Dataset -m ./pretrainedModel -l ./valid -s valid

and then i used this command to evaluate

python3 ./utils/evaluate_mos.py -d Dataset -p ./valid --datacfg ./config/labels/semantic-kitti-mos.yaml

[MaxChanger]

Ok, I'm going to retest, but since my gpu server crashed the other day, I need to reconfigure the environment. please hold on

[MaxChanger]

Hi, sorry for the long wait, I found out that the problem is because nan appears during inference.
This problem is also mentioned in #9 and #6, I will debug why this happens

[MaxChanger]

Hi @Terminal-K , I finally found the cause of this problem!
I ignored the version of softpool, its author updated the repo on 2022/04/07, and our project was implemented before that, using its historical version, the commits id is 2d2ec6d.

When I use the new version d056ab8 of softpool code for inference, nan will also appear, just like what you encounter in the process of inference or train.

The specific reason may need to carefully check the softpool code and discuss with the author, but rolling back the softpool version is a quick solution to this problem.

git clone https://github.com/alexandrosstergiou/SoftPool.git
cd SoftPool
git checkout 2d2ec6d # rollback to 2d2ec6dca10b7683ffd41061a27910d67816bfa5
cd pytorch
make install
--- (optional) ---
make test

I hope you can help to check it, if you have any questions, please contact me again.

[Terminal-K]

Thank you for your patience, i will try again as you say:)