nnunet pkl文件

[Sere1nz]

问题确认 Search before asking

• [X] 我已经搜索过问题，但是没有找到解答。I have searched the question and found no related answer.

请提出你的问题 Please ask your question

你好，如果我使用nnunet只想进行肺结节推理的话，我看需要一个flag是plan_path，这个pkl文件应该在哪下载或者生成，还是说直接忽略这个flag就好？权重文件我已经下好了。

[weihuang-cs]

plans.pkl这个文件基本参照了nnunet原始的设定，区别在于paddleseg把他放到了datasets初始化的时候 下面是一个示例：

%cd ~/PaddleSeg/contrib/MedicalSeg/

from nnunet.datasets import *
config = dict(
  plans_name= 'nnUNetPlansv2.1_plans_3D.pkl',
  dataset_root= '/',
  result_dir= '/',
  raw_data_dir= 'msd_lung/Task06_Lung',
  decathlon_dir= 'msd_lung/Task006_Lung',
  cropped_data_dir= 'msd_lung/cropped/Task006_Lung',
  preprocessed_dir= 'msd_lung/preprocessed/Task006_Lung',
  plan2d= True,
  plan3d= False,
  num_batches_per_epoch= 250,
  fold= 2,
  stage= 1,
  unpack_data= True,
  cascade= True,
  mode= 'train',
)
dataset = MSDDataset(**config)

读取plan文件：

import pickle
def load_pickle(file):
    with open(file, 'rb') as f:
        data = pickle.load(f)
    return data
load_pickle("msd_lung/preprocessed/Task006_Lung/nnUNetPlansv2.1_plans_2D.pkl")

输出结果：

{'num_stages': 1,
 'num_modalities': 1,
 'modalities': {0: 'CT'},
 'normalization_schemes': OrderedDict([(0, 'CT')]),
 'dataset_properties': {'all_sizes': [(304, 512, 512)],
  'all_spacings': [array([1.        , 0.69335938, 0.69335938])],
  'all_classes': [1],
  'modalities': {0: 'CT'},
  'intensityproperties': OrderedDict([(0,
                OrderedDict([('local_props',
                              OrderedDict([('lung_001',
                                            OrderedDict([('median', -42.0),
                                                         ('mean', -144.93015),
                                                         ('sd', 201.28882),
                                                         ('mn', -764.0),
                                                         ('mx', 87.0),
                                                         ('percentile_99_5',
                                                          84.86999999999989),
                                                         ('percentile_00_5',
                                                          -682.74)]))])),
                             ('median', -42.0),
                             ('mean', -144.93015),
                             ('sd', 201.28882),
                             ('mn', -764.0),
                             ('mx', 87.0),
                             ('percentile_99_5', 84.86999999999989),
                             ('percentile_00_5', -682.74)]))]),
  'size_reductions': OrderedDict([('lung_001', 1.0)])},
 'list_of_npz_files': ['msd_lung/cropped/Task006_Lung/lung_001.npz'],
 'original_spacings': [array([1.        , 0.69335938, 0.69335938])],
 'original_sizes': [(304, 512, 512)],
 'preprocessed_data_folder': 'msd_lung/preprocessed/Task006_Lung',
 'num_classes': 1,
 'all_classes': [1],
 'base_num_features': 32,
 'use_mask_for_norm': OrderedDict([(0, False)]),
 'keep_only_largest_region': None,
 'min_region_size_per_class': None,
 'min_size_per_class': None,
 'transpose_forward': [0, 1, 2],
 'transpose_backward': [0, 1, 2],
 'data_identifier': 'nnUNetData_plans_v2.1_2D',
 'plans_per_stage': {0: {'batch_size': 12,
   'num_pool_per_axis': [7, 7],
   'patch_size': array([512, 512]),
   'median_patient_size_in_voxels': array([304, 512, 512]),
   'current_spacing': array([1.        , 0.69335938, 0.69335938]),
   'original_spacing': array([1.        , 0.69335938, 0.69335938]),
   'pool_op_kernel_sizes': [[2, 2],
    [2, 2],
    [2, 2],
    [2, 2],
    [2, 2],
    [2, 2],
    [2, 2]],
   'conv_kernel_sizes': [[3, 3],
    [3, 3],
    [3, 3],
    [3, 3],
    [3, 3],
    [3, 3],
    [3, 3],
    [3, 3]],
   'do_dummy_2D_data_aug': False}},
 'preprocessor_name': 'PreprocessorFor2D'}