About "train_fine_18unlabel_pseudo_TTA_1t_deeplabv3.txt" file
daeunni opened this issue · comments
Daeun Lee commented
Hi,
Thanks for your nice work.
Btw, I'm trying to run this code with cityscapes dataset.
According to your sample .yaml file, cityscapes uses train_18unlabel_deeplabv3_1t
split. (like below)
DATASET:
NAME: "cityscape"
MEAN: [0.5, 0.5, 0.5]
STD: [0.5, 0.5, 0.5]
MODE: 'testval'
SEMI:
TRAIN_SPLIT_L: 'train_18unlabel_deeplabv3_1t' # split
TRAIN_SPLIT_U: 'train_18unlabel_deeplabv3_1t' # split
VAL_SPLIT: 'val_fine'
However, in the SimpleBaseline/segmentron/data/dataloader/cityscapes.py
L42 we need train_fine_18unlabel_pseudo_TTA_1t_deeplabv3.txt file.
elif split == 'train_18unlabel_deeplabv3_1t':
_split_f = [
"data/train_fine_18unlabel_pseudo_TTA_1t_deeplabv3.txt",
"data/train_fine_18label.txt"]
Where can I find this file?
Daeun Lee commented
Or can I replace this part like below?
I want to set semi-supervised setting.
DATASET:
NAME: "cityscape"
MEAN: [0.5, 0.5, 0.5]
STD: [0.5, 0.5, 0.5]
MODE: 'testval'
SEMI:
# DE :: modified
TRAIN_SPLIT_L: 'train_18label'
TRAIN_SPLIT_U: 'train_18unlabel'
VAL_SPLIT: 'val_fine'