MedSegDiff

MedSegDiff copied to clipboard

The poor performances of the multi-labels BraTs dataset

8

The raw images:     The Ground Truth:  The output sample of 5 ensembles:  So why these outputs contain the brain boundary? And...

smallboy-code

pretrained model of diffusion model

1

Thank you for your excellent work！ Is there a pretrained model of diffusion model that can be provided at training time.

JAYCHOU2020

Too many iterations lead to rapid performance degradation

2

I trained the model on the ISIC dataset with the default experimental settings provided in the readme and sampled the same image using the weight files of 15,000, 65,000 and...

RunpuWei

about the code in p_sample_loop_known

2

if dice_score(final["sample"][:,-1,:,:].unsqueeze(1), final["cal"]) < 0.65: cal_out = torch.clamp(final["cal"] + 0.25 * final["sample"][:,-1,:,:].unsqueeze(1), else: cal_out = torch.clamp(final["cal"] * 0.5 + 0.5 * final["sample"][:,-1,:,:].unsqueeze(1), 0, 1) What is the main purpose of...

jaceqin

BRATS Dataset training testing split

1

Hi there, nice work. Can you provide me your training and testing split for the BRATS21 dataset? I am trying to reproduce your work so I would like to know...

rtx98

use DDIM error

5

xupinggl

After setting use_fp16 = True, some bugs occur:  It seems that not all of the models and weights are converted to fp16? Looking forward to your reply, thank you

RoboticsZhang

关于dpm-solver的问题

3

作者你好，我在训练ISIC数据集的时候，训练了2w7个step，使用ddim sample可以得到较好的结果，但是当我使用dpm-solver之后效果会变得很差，请问这是什么原因呢？ dpm-solver的配置是50个step。 期待您的回复

hongkaisdfs

Some questions about DPM-solver

2

Hi Great job! I have some question when i used "DPM-solver=True". 1: Does the batchsize must be 1 when used DPM-solver ? if not in dpm_solver.py def dynamic_thresholding_fn(self, x0, t):...

LiuTingWed

 Thank you for your great job. I am relly interesting to your job V1 and V2. Could you please explain the mean of larger variance generated by MedSegDiff-V1. Why...

yoookoso