关于AIIB2023数据集的处理

[wakakaFF]

你好，前辈。
我想请教一下，如何用你的训练框架处理AIIB2023数据集。
我尝试性的试了一下。以下是我的处理以及报错详情。

[ge-xing]

你代码 有点问题，用我这个

from light_training.preprocessing.preprocessors.default_preprocessor import DefaultPreprocessor 
import numpy as np 
import pickle 
import json 

base_dir = "/data/xingzhaohu/aiib23/data/raw_data/AIIB23_Train_T1"
image_dir = "img"
gt_dir = "gt"

def process_train():
    # fullres spacing is [0.5        0.70410156 0.70410156]
    # median_shape is [602.5 516.5 516.5]
    preprocessor = DefaultPreprocessor(base_dir=base_dir, 
                                    image_dir=image_dir,
                                    label_dir=gt_dir,
                                   )

    out_spacing = [0.5, 0.70410156, 0.70410156]
    output_dir = "./data/fullres/train/"
    
    with open("./data_analysis_result.txt", "r") as f:
        content = f.read().strip("\n")
        print(content)
    content = eval(content)
    foreground_intensity_properties_per_channel = content["intensity_statistics_per_channel"]

    preprocessor.run(output_spacing=out_spacing, 
                     output_dir=output_dir, 
                     all_labels=[1],
                     foreground_intensity_properties_per_channel=foreground_intensity_properties_per_channel
    )

def plan():
    
    preprocessor = DefaultPreprocessor(base_dir=base_dir, 
                                    image_dir=image_dir,
                                    label_dir=gt_dir,
                                   )
    preprocessor.run_plan()


if __name__ == "__main__":
# 
    # plan()

    process_train()
    # process_val()
    # process_test()
    

[ge-xing]

你先运行plan函数会在当前目录下得到 data_analysis_result.txt 第二步预处理 会用到这个文件。

[wakakaFF]

你先运行plan函数会在当前目录下得到 data_analysis_result.txt 第二步预处理 会用到这个文件。

非常感谢。