Using Models from the Zoo on my custom datasets

[AndromathArcanitus]

Dear All,

I am new to MONAI and am hoping to use the model lung nodule ct detection from Model Zoo on my own dataset. I have already implemented my DataLoader. So, when I call it, the data is ready. But, when I see the configuration json file, the data is load through

"data_list_file_path": "$@bundle_root + '/LUNA16_datasplit/dataset_fold0.json'",
"dataset_dir": "/datasets/LUNA16_Images_resample",
"train_datalist": "$monai.data.load_decathlon_datalist(@data_list_file_path, is_segmentation=True, data_list_key='training', base_dir=@dataset_dir)",.

Can anyone provide any suggestions for my case? Should I save the data into a folder? But I don't know in what form the data should be saved. Thank you.

[KumoLiu]

Hi @AndromathArcanitus, you can change the data path in the configure file if you want to run bundle. Otherwise you could directly using the model weight in your pipeline. Thanks!

[AndromathArcanitus]

Hi @KumoLiu , Thank you very much for the timely response. I will certainly try it.

[AndromathArcanitus]

Hi @KumoLiu, since I processed my data using different approaches from those for LUNA16, I have already written my DataLoader and loaded the data into the memory (I guess), using, for example,

train_dataset = LungNoduleDataModule(data_folder=data_folder).train_dataloader()

Can I ask how to guide the model to read my data? Thank you.

[KumoLiu]

Hi @AndromathArcanitus, if you want to train your pipeline in a pythonic way, you can just follow this tutorial which explore how to access a bundle in Python and use it in your own application.

BTW, you can also find more use case about how to use the bundle in MONAI Zoo here.

Hope it helps, thanks!

[AndromathArcanitus]

It does. Thank you very much.