search

ubc-tea / SADM-Longitudinal-Medical-Image-Generation

The official implementation of our IPMI 2023 paper "SADM: Sequence-Aware Diffusion Model for Longitudinal Medical Image Generation"
MIT License
51 stars 3 forks source link

It's weird that I ended up with a bunch of mosaics #1

Open guanzhenghua opened 1 year ago

guanzhenghua commented 1 year ago

I use the following data set in the code, first run ACDC_loader, then ACDC_prapare, and finally run SADM, leaving the default parameters, and finally loss is 0.1691, visualized as a bunch of mosaics. Thank you very much for the answer. This is very important to me. You can also contact me by email. My email address is guanzhenghua2022@email.szu.edu.cn

wltjr1007 commented 1 year ago

Current commit contains the minimal code for training and validation (i.e., hyperparameters are not the same as experiments in the paper). The full code will be ready before the camera-ready version of the paper is published (or hopefully sooner). Please stay tuned.

guanzhenghua commented 1 year ago

Thank you very much for your answer. Because of current commit only contains the minimal code for training and validation. I am worried about whether the results similar to those in the paper can be obtained just by adjusting hyperparameters. About using the dataset in the code. Could you give me some hints about hyperparameters. My gpu is Nvidia A6000 48GB. Thanks a lot.

Barrett-python commented 1 year ago

I have the same problem when I run the code.