Config file for training luna_ncc from scratch

EndoluminalSurgicalVision-IMR / Medical-SSL

[MedIA 2023]The official implementation of MIA paper "Dive into the details of Self-supervised Learning for Medical Image Analysis"

28 stars 3 forks source link

Config file for training luna_ncc from scratch #3

Open felicitas-b opened 5 months ago

felicitas-b commented 5 months ago

Hi, thank you for your work! We tried to reproduce your results for the NCC on LUNA2016 from scratch (see table 4 row 1 in your paper) with the provided config luna_ncc_3d_config.py but were not able to match them. Could you please provide us with the configurations you used for training the networks from scratch?

Thanks in advance!

chuyan99 commented 5 months ago

If you set " pretrained_model = None" in the config.py, the model would be trained from scratch. The other settings are the same as the provided config luna_ncc_3d_config.py.