Hello,
Any update about docker submission?
What will be the new deadline for final submission?
Thanks.
Best regards,
Abdul
Created by Abdul Qayyum abdulkcl Further extension: 2 days (11.04.2023 23:59 UTC)
@SaskiaB : one text file with all the info is enough. Hi,
If we want to submit several containers via docker hub, do we need to upload text files for each method individually? Or can we just include the links to all containers in one txt file?
Thanks,
Saskia Hello @soumick
Thank you for the precision! Hey @aryapa
You can build the container, test it, and then save it using docker save (cf. https://docs.docker.com/engine/reference/commandline/save/).
After you build the tar, you can upload it to our cloud.
Or else, you can also push it to the Docker Hub and share with us the link (upload a txt file with your team name to our cloud) Thank you for giving more insights.
I am used to pulling and pushing dockers but have never dragged and dropped one onto a cloud could you explain how we are supposed to push it to this website ? ( https://cloud.ovgu.de/s/QyMZK9nJJscDk3c )
Or do you want a zip file with the docker file the models and the eval.py and you build the container yourself ?
Best, Instructions for Docker Creation
You need to submit a docker container for evaluation. Internally, the structure of the container is up to you, except for a few requirements. The container must contain an "eval.py" script that accepts "ds_path" and "out_path" as keyworded command line arguments. ds_path will be the path to our dataset that we will provide during inference. Inside the provided path, there will be a directory named "test" - containing all the test volumes in ".nii.gz" format. They are all 3D 7T MRA volumes similar to the ones made public for training. The code should be able to handle any arbitrary shapes of input volumes. You can pad, crop or modify them as required by your algorithm. The system where the executions will be performed is equipped with a CUDA-enabled GPU with 48GB memory (Nvidia A6000), a 16-core 32-thread processor (AMD Ryzen 9 3950X) and 64 GB of RAM. Your program must be able to run on this system. Any docker which fails to run (e.g. results in CUDA out of memory or hangs the system due to over-use of CPU or RAM) will be disqualified. There will not be any internet access provided. Hence, the docker container needs to be self-contained (i.e. any trained weights, including any required pre-trained weights, must be present inside the container). After execution, your program must store the produced output inside the specified out_path in ".nii.gz" format with the same name as the input volume (i.e. if there was a file named spock.nii.gz inside the ds_path, the output of your program should also be called spock.nii.gz and stored inside the out_path) inside a new folder with your team name. After execution, if the results are produced in the appropriate directory, the submissions will be considered valid.
Apart from the "eval.py", your container should also contain one more file - "info.txt". This file should contain the name of your team, the list of your team members and their affiliations, and the title of your abstract. The abstract itself should not be here.
The size of the docker container should be less than 10 GB. If your container is larger than that, please email us before submitting it.
Instructions for Docker Submission
You can upload your docker containers to our University cloud: https://cloud.ovgu.de/s/QyMZK9nJJscDk3c with your team name as the name of the container.
If the name of the file is anything other than the team name, it won't be considered.
Each team can upload multiple dockers for multiple methods. In that case, you need to add an acronym of your method to your team name. (e.g. picard_phaser and picard_tricorder for phase and tricorder methods by the team picard). Only one submission per team per method will be considered. But you are allowed to upload multiple times - only the latest one will be considered. While storing the outputs, they should also use this same naming convention (i.e. "teamname_methodacronym" if multiple methods, "teamname "if only one method). For any query or assistance required regarding the creation of the docker or the uploading process, please contact Mr Hannes Schnurre (email: hannes.schnurre@ovgu.de OR phone: +49 391 67-56116).
Instructions for uploading the abstract
You can upload your abstract to our University cloud: https://cloud.ovgu.de/s/QyMZK9nJJscDk3c with your team name as the name of the files. You need to upload 2 files: a PDF and a DOC or LaTeX. You can follow any template, but we recommend using the ISBI template. Both the PDF and the editable file (in doc or docx or tex format) should be with your team name (or teamname_methodname if submitting multiple methods).
Challenge Timeline
Docker and Abstract Submission: OPEN NOW
Docker and Abstract Submission Deadline #1: 09.04.2023 23:59 UTC (to be included in the challenge at ISBI 2023, as well as on the challenge manuscript)
Docker and Abstract Submission Deadline #2: 23.04.2023 23:59 UTC (won't be declared as a winner even if it's the best performing method, to be included only on the challenge manuscript).
For any technical support or queries, please contact Mr Hannes Schnurre (email: hannes.schnurre@ovgu.de OR phone: +49 391 67-56116).
For any other queries, feel free to contact us directly. Dear Organizers,
Our team is also very eager to participate and we are quite unsure about the format of the docker submission you are expecting. Is it of a similar nature to the SHINY-ICARUS challenge ?
What is the submission deadline and could you provide guidelines as to how you wish us to submit our dockers and abstracts.
Best,
Arya.