轻松nnDetection自己的训练集

大神就是大神，一定要紧跟步伐。
Michael Baumgartner在nnUnet之后推出了nnDetection，仔细一看，居然三年前就已经推出了medicaldetectiontoolkit，里面集成了Mask R-CNN, Faster R-CNN+ 等。
追随最新代码有网站https://paperswithcode.com/。

论文地址：https://arxiv.org/pdf/2106.00817v1.pdf
代码地址：https://github.com/MIC-DKFZ/nnDetection

下载源代码解压缩

Dockerfile代码稍作修改：

#Copyright 2020 Division of Medical Image Computing, German Cancer Research Center (DKFZ), Heidelberg, Germany
#
#Licensed under the Apache License, Version 2.0 (the "License");
#you may not use this file except in compliance with the License.
#You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
#Unless required by applicable law or agreed to in writing, software
#distributed under the License is distributed on an "AS IS" BASIS,
#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
#See the License for the specific language governing permissions and
#limitations under the License.

# Contains pytorch, torchvision, cuda, cudnn
FROM nvcr.io/nvidia/pytorch:20.12-py3

ARG env_det_num_threads=6
ARG env_det_verbose=1

# Setup environment variables
ENV det_data=/opt/data det_models=/opt/models det_num_threads=$env_det_num_threads det_verbose=$env_det_verbose OMP_NUM_THREADS=1

# Install some tools
RUN apt-get update && export DEBIAN_FRONTEND=noninteractive && apt-get install -y \
 git \
 cmake \
 make \
 wget \
 gnupg \
 build-essential \
 software-properties-common \
 gdb \
 ninja-build

RUN pip install pip -U && pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple

RUN pip install numpy

# Install own code
COPY ./requirements.txt .
RUN mkdir ${det_data} \
  && mkdir ${det_models} \
  && mkdir -p /opt/code/nndet \
  && pip install -r requirements.txt  \
  && pip install hydra-core --upgrade --pre \
  && pip install pytorch-model-summary   
#  && pip install git+https://github.com/mibaumgartner/pytorch_model_summary.git

WORKDIR /opt/code/nndet
COPY . .
RUN FORCE_CUDA=1 pip install -v -e .

docker build -t nndetection:0.1 --build-arg env_det_num_threads=6 --build-arg env_det_verbose=1 .

docker run --gpus all -v ${det_data}:/opt/data -v ${det_models}:/opt/models -it --shm-size=24gb nndetection:0.1 /bin/bash

docker run --gpus all -v /home/yakeworld/work/nnDetection/data:/opt/data -v /home/yakeworld/work/nnDetection/models:/opt/models -it --shm-size=24gb nndetection:0.1 /bin/bash

docker run --gpus all -v /home/amax/work/nnDetection/data:/opt/data -v /home/amax/work/nnDetection/models:/opt/models -it --shm-size=24gb nndetection:0.1 /bin/bash




docker run --gpus all  --ipc=host -it --rm  --shm-size=24gb nndetection:0.1 /bin/bash

docker run --gpus all -v /home/yakeworld/datas:/opt/data -v /home/yakeworld/models:/opt/models -it --shm-size=24gb nndetection:0.1 /bin/bash

docker run  -d --name nndetection  --gpus all -v /home/yakeworld/work/data/:/opt/data -v /home/yakeworld/work/models:/opt/models -it --shm-size=24gb nndetection:0.1 /bin/bash

nndet_example 可以生产测试集，观察相应目录和文件结构。

依样画葫芦，创建自己数据的相关文件结构。

data.json

{
    "task": "Task04_Hippocampus",
    "name": "Hippocampus",
    "target_class": null,
    "test_labels": true,
    "labels": {
        "0": "background",
        "1": "Anterior", 
        "2": "Posterior"
                },
    "modalities": {
        "0": "MRI"
    },
    "dim": 3
}
 

tag.json

{
    "instances": {
        "1": 1,
        "2": 2
    }
}

nndet_prep 04

nndet_unpack preprocessed/D3V001_3d/imagesTr 6

nndet_train 04
nndet_eval 091 RetinaUNetV001_D3V001_3d 0 --boxes --analyze_boxes

nndet_consolidate 091 RetinaUNetV001_D3V001_3d --sweep_boxes

nndet_predict 04 RetinaUNetV001_D3V001_3d --fold -1

半规管数据集

拷贝nnUnet生成的目录
1.数据预处理
nndet_prep 091
2.数据解压缩
cd /opt/data/Task091_innerear
nndet_unpack preprocessed/D3V001_3d/imagesTr 6
3.训练模型
nndet_train 091

https://github.com/GJiananChen/MICCAI2021-OpenReviewAnalysis#opensource

仔细阅读论文，分析思路，是非常有效的学习方法。
最好是具备一定的重现代码能力。

nnDetection利用了Retina U-Net作为基础，这个一个比较创新的模型，需要仔细学习。

https://blog.csdn.net/qq_41084756/article/details/96735852