Dataloading process

[guangyaoz]

Hello,

I am trying to reproduce your experiment on my local machine, but I am not sure how you loaded the data. What should the training testing and validation data directory look like? Thanks.

[JayParanjape]

Hi,
The data folder can be anything that makes it easy for you to add the corresponding code to data_utils.py.
For example, you can create a class called Custom_Dataset in data_utils.py. This class should have a get_item function that returns img, mask, path of the image and the labelname. As an example you can refer to any one of the datasets, like class ChestXDet

[JayParanjape]

If you can modify the data structure available to you, I would recommend a root directory that contains the folders train, test and val as follows:
root
| - train
|- images
|- masks
| - val
|- images
|- masks
| - test
|- images
|- masks
Then, you can load from these folders in the populate lists function in data_utils.py

[guangyaoz]

Thanks!