med seg diff pytorch

MedSegDiff - Pytorch

在 Pytorch 中实现 MedSegDiff - 百度使用 DDPM 进行 SOTA 医学分割，并在特征级别上增强调节，并在傅立叶空间中过滤特征。

欣赏

• StabilityAI 以及我的其他赞助商的慷慨赞助

• Isamu 和 Daniel 为皮肤病变数据集添加了训练脚本！

安装

$ pip install med-seg-diff-pytorch

用法

 import torch
from med_seg_diff_pytorch import Unet , MedSegDiff

model = Unet (
    dim = 64 ,
    image_size = 128 ,
    mask_channels = 1 ,          # segmentation has 1 channel
    input_img_channels = 3 ,     # input images have 3 channels
    dim_mults = ( 1 , 2 , 4 , 8 )
)

diffusion = MedSegDiff (
    model ,
    timesteps = 1000
). cuda ()

segmented_imgs = torch . rand ( 8 , 1 , 128 , 128 )  # inputs are normalized from 0 to 1
input_imgs = torch . rand ( 8 , 3 , 128 , 128 )

loss = diffusion ( segmented_imgs , input_imgs )
loss . backward ()

# after a lot of training

pred = diffusion . sample ( input_imgs )     # pass in your unsegmented images
pred . shape                              # predicted segmented images - (8, 3, 128, 128) 

训练

运行命令

accelerate launch driver.py --mask_channels=1 --input_img_channels=3 --image_size=64 --data_path= ' ./data ' --dim=64 --epochs=100 --batch_size=1 --scale_lr --gradient_accumulation_steps=4

如果你想添加我们用目前为止拥有的掩码进行调节的自我条件，请执行 --self_condition

托多

• 一些基本的训练代码，训练器采用针对医学图像格式定制的自定义数据集 - 感谢@isamu-isozaki
• 中间任意深度的完整变压器，如简单扩散中所做的那样

引文

 @article { Wu2022MedSegDiffMI ,
    title   = { MedSegDiff: Medical Image Segmentation with Diffusion Probabilistic Model } ,
    author  = { Junde Wu and Huihui Fang and Yu Zhang and Yehui Yang and Yanwu Xu } ,
    journal = { ArXiv } ,
    year    = { 2022 } ,
    volume  = { abs/2211.00611 }
}

 @inproceedings { Hoogeboom2023simpleDE ,
    title   = { simple diffusion: End-to-end diffusion for high resolution images } ,
    author  = { Emiel Hoogeboom and Jonathan Heek and Tim Salimans } ,
    year    = { 2023 }
}