KeyError Problem

[ZhouL2001]

Dear author, thank you for your good work. I tried to re-train your network, but raise a key error:
KeyError: "Dataset 'polyp_hard_unseen_validate' is not registered! Available datasets are: coco_2014_train,...
Can you provide some solutions to solve the problem?
Thank you very much.

[bio-mlhui]

Have you set the env variable export CURRENT_TASK=VIS (Video Instance Segmentation) before executing main? This line

LGRNet/data_schedule/__init__.py

Line 4 in 8672bcf

if os.getenv('CURRENT_TASK') == 'VIS':

should import the "polyp_hard_unseen_validate" dataset.

Feel free to let me know if you have other questions!

[ZhouL2001]

Thank you very much. If I want to train this model on my own datasets, can you provide some instructions.

[bio-mlhui]

Assume you are doing video semantic segmentation.

1. Make sure the data storage format be the same as:

-- Frame
    -- vid1
        -- img file
-- GT
    -- vid1
        -- mask file

as introduced in https://github.com/bio-mlhui/LGRNet/blob/main/assets/DATA.md

2. register this dataset under data_schedule/vis/ similar to https://github.com/bio-mlhui/LGRNet/blob/main/data_schedule/vis/polyp/polyp_dataset.py