Closed yyf17 closed 1 year ago
The setup is straightforward - there is a folder for masks that contains the mask files, and another folder for images that stores the corresponding RGB images (it is important to verify the file names in the dataloader prior to training). The matching of masks and RGB images is based on their shared file names, which are usually numbered from 1 to N, where N represents the total number of images.
Very wonderful work! Can you provide the structure of the dataset?