Timeout msn()

[jdhenaos]

Hi!

I was trying to use the ASOCA dataset. I started with the first code lines in the GetStarted notebook:

!pip install MedShapeNet

from MedShapeNet import MedShapeNet as msn
msn_instance = msn(timeout=120)

However, I get a timeout error even when increasing the time parameter:

Connection to MinIO server timed out after 120 seconds.
Download directory already exists at: /content/msn_downloads

Do you know if there is a previous step I have omitted? How can I ensure the connection to the server?

Thanks in advance

[GLARKI]

Dear Jdhenaos,

First of all, thank you for emailing me! You did everything correctly, let me explain what happened, how it will be prevented in the future and how you can still use the api.

What happened:

• the MinIO server runs on a windows computer and during my trip to MICCAI in Marrakesh, windows decided it was time for a forced update and restart, thus docker desktop and the DUC wasn't running to handle the dynamic dns. Therefore you could haven ot reached the server over this port 9000.

Future workaround

• currently I have the MedShapeNetCore dataset (a selection of downsized datasets) on my MinIO storage so that users and contributors can already check it out and develop. I'm currently filing the paperwork to get S3 storage on Coscine https://about.coscine.de/en/ -> I expect to have this within 3-4 months max. Then there will be practically zero downtime, increased download speed and 100tb of storage so that I can make all datasets available in the same format as MedShapeNetCore is currently.

What you can still do:

• So data from the MinIO bucket i.e. MedShapeNetCore is not available and all functions that try to use this. However some functions convert stl to ply or binary masks, these are still available. MORE IMPORTANTLY all shapes on the current slow Sciebo storage are still available. These are over 100k shapes. You can use the search_by_name and get lists of files/urls for these shapes. You can also use the search_and_download_by_name function and download them directly to create your own datasets based on shape names. You can checkoud the beginning of the pointcloudcompletor example for this.

Again thank you very much for notifying me of this downtime <3, please stay tuned for when we have the entire dataset on COSCINE and make MedShapeNet 2.0!

Feel free to ask me anything, anytime.