By: ashkanpakzad on Aug. 16, 2022, 3:57 p.m.
Dear ATM organisers,
I note the test data set is 150 images and that the docker submission will only be allowed to execute for 6 hours. This allows about 2.4 minutes per case.
This to me seems a relatively short amount of time for a 3D medical image segmentation task that doesn't focus on speed. I just wanted to confirm, is 6 hours the final designated time for the full 150 test cases?
Many thanks, Ashkan
By: minghuizhang24 on Aug. 17, 2022, 4:45 a.m.
Hi,
Thank you for your question. We have tested our baseline and 6 hours are enough for 150 images prediciton. However, consider the participants' request, we extend the docker running hours to 12 hours preferably. We modify the docker running time rule to : "The docker should execute preferably no more than 6 12 hours and occupy preferably no more than 16 GB GPU Memory to generate the segmentation test results. " In addition, we will send all of the participants an official e-mail about the test phase on 18/08/2022.
However, we will calculate the inference time and GPU memory as efficiency reference, which will be reported on our challenge satellite event, 22/09/2022.
We will take care of your submitted docker, and maybe the efficiency of the airway tree modeling is one of the future direction we wanna explore.
Best Regards, On the behalf of Institute of Medical Robotics, Shanghai Jiao Tong University
