How to reproduce using own dataset

[JFJ-Bin]

I'm going to use the author's code to replicate my own dataset, can you tell me how to generate the training.npy file in the image below?

Or can you say how to modify it to train own dataset? Thank you!

[QLYYLQ]

as shown in code: dataset/voc.py. You should set a own Segmentation(like VOCSegmentation in voc.py) which can use getitem to visit total dataset, then you can write an incremental dataset, like VOCSegmentationIncremental. I write a base datasets for those in my repo: When sam meets continual learning.