CUDA_VISIBLE_DEVICES=0 python train_search.py when we run this command getting error?

[Naveennavik]

Your work seem to be good enough to see but i have some doubts When i tried running the above command it showing error Traceback (most recent call last): File "train_search.py", line 23, in from datasets.cityscapes.cityscapes import Cityscapes File "/media/drive/2f08d5f1-1439-4edf-9bb4-1d19896475a5/Naveen/FasterSeg/tools/datasets/init.py", line 1, in from .cityscapes import Cityscapes File "/media/drive/2f08d5f1-1439-4edf-9bb4-1d19896475a5/Naveen/FasterSeg/tools/datasets/cityscapes/init.py", line 1, in from .cityscapes import Cityscapes File "/media/drive/2f08d5f1-1439-4edf-9bb4-1d19896475a5/Naveen/FasterSeg/tools/datasets/cityscapes/cityscapes.py", line 6, in class Cityscapes(BaseDataset): TypeError: module.init() takes at most 2 arguments (3 given) May i know what is the reason for this error @chenwydj @TianlongChenTAMU

[chenwydj]

Hi @Naveennavik! Thank you for your interest in our work!

I just cloned my repo and tested it again, and the search script works. I did not meet your error.

Could you please make sure you are on the latest commit?

[Naveennavik]

Thanks for replying,I'm using latest commit only . Can i use the test to work on my own dataset?

[chenwydj]

You can learn how to write your own dataset by following BaseDataset.py and cityscapes.py.

[Naveennavik]

In BaseDataset.py main these are the line i need to give path right? img_root=is the path to the image folder whereas train_source and eval_source is what may i know Big Thanks in advance data_setting = {'img_root': '', 'gt_root': '', 'train_source': '', 'eval_source': ''}

[chenwydj]

Please follow this configuration file.

[Naveennavik]

For the new dataset we need train and then only we can test right orelse we can directly run test code? I dont understand clearly on the documentation how to run ?

[chenwydj]

If your dataset is very different from Cityscapes (e.g. different number of semantic classes, large appearance gap), then yes, you first need to train on your dataset, then evaluate.

[Naveennavik]

cityscapes_train_fine.txt cityscapes_train_val_fine.txt cityscapes_val_fine.txt cityscapes_test.txt How to prepare these txt files for further process

[chenwydj]

You can just use python to list all the file names and write them into a file with the same format.