search

mattiaspaul / OBELISK

MIDL 2018 / MEDIA 2019: one binary extremely large and inflecting sparse kernel (pytorch)
MIT License
44 stars 9 forks source link

Clarification #1

Open amirid opened 5 years ago

amirid commented 5 years ago

Hi Mattias,

What does 9 indicate in the following line? https://github.com/mattiaspaul/OBELISK/blob/c757ebae7a84723ff6109c375771f1cb3e5379ca/midl_script_augment_github.py#L142

Thanks.

mattiaspaul commented 5 years ago

It indicates that (in my case) the size of the training database contains 9 (3D) scans. The next version (which I plan to upload by the end of the year) will contain a more flexible way to train with custom datasets.

Dongshengjiang commented 5 years ago

I would like to say it is another good work from you, I am a follower of you from self-similarity feature to CNN. The idea of using a large and dynamic kernel would be great innovation. Waiting and hoping you could update with the full source code with pretrained model soon. Cheers, Mattias!