Layout training directory?

[SjoerdTimovanVliet]

Dear Sir/Madam,

I would like to train your model on my own dataset. What is the layout of the directory with groundtruths and images such that I can train my model?

[th2l]

Hi, The _rootpath (line 322,327,331) in train.py should have the structure as below

root_path/
├─ train/
│  ├─ images/
│  │  ├─ imgtr_00001.png
│  │  ├─ ....
│  ├─ labels/
│  │  ├─ imgtr_00001.npy
│  │  ├─ .....
├─ validation/
│  ├─ images/
│  │  ├─ imgval_00001.png
│  │  ├─ .....
│  ├─ labels/
│  │  ├─ .....
│  │  ├─ imgval_00001.npy
├─ test/
│  ├─ images/
│  │  ├─ imgtest_00001.png

It similar to extracted folder of zip file from Semantic_Segmentation_Dataset.zip

You can also modify above 3 lines, and OpenEDS class (utils.py, line 91, 111) to your own folder structure.