Spaces:
Sleeping
Sleeping
""" | |
This file records the directory paths to the different datasets. | |
You will need to configure it for training the model. | |
All datasets follows the following format, where fgr and pha points to directory that contains jpg or png. | |
Inside the directory could be any nested formats, but fgr and pha structure must match. You can add your own | |
dataset to the list as long as it follows the format. 'fgr' should point to foreground images with RGB channels, | |
'pha' should point to alpha images with only 1 grey channel. | |
{ | |
'YOUR_DATASET': { | |
'train': { | |
'fgr': 'PATH_TO_IMAGES_DIR', | |
'pha': 'PATH_TO_IMAGES_DIR', | |
}, | |
'valid': { | |
'fgr': 'PATH_TO_IMAGES_DIR', | |
'pha': 'PATH_TO_IMAGES_DIR', | |
} | |
} | |
} | |
""" | |
DATA_PATH = { | |
'videomatte240k': { | |
'train': { | |
'fgr': 'PATH_TO_IMAGES_DIR', | |
'pha': 'PATH_TO_IMAGES_DIR' | |
}, | |
'valid': { | |
'fgr': 'PATH_TO_IMAGES_DIR', | |
'pha': 'PATH_TO_IMAGES_DIR' | |
} | |
}, | |
'photomatte13k': { | |
'train': { | |
'fgr': 'PATH_TO_IMAGES_DIR', | |
'pha': 'PATH_TO_IMAGES_DIR' | |
}, | |
'valid': { | |
'fgr': 'PATH_TO_IMAGES_DIR', | |
'pha': 'PATH_TO_IMAGES_DIR' | |
} | |
}, | |
'distinction': { | |
'train': { | |
'fgr': 'PATH_TO_IMAGES_DIR', | |
'pha': 'PATH_TO_IMAGES_DIR', | |
}, | |
'valid': { | |
'fgr': 'PATH_TO_IMAGES_DIR', | |
'pha': 'PATH_TO_IMAGES_DIR' | |
}, | |
}, | |
'adobe': { | |
'train': { | |
'fgr': 'PATH_TO_IMAGES_DIR', | |
'pha': 'PATH_TO_IMAGES_DIR', | |
}, | |
'valid': { | |
'fgr': 'PATH_TO_IMAGES_DIR', | |
'pha': 'PATH_TO_IMAGES_DIR' | |
}, | |
}, | |
'backgrounds': { | |
'train': 'PATH_TO_IMAGES_DIR', | |
'valid': 'PATH_TO_IMAGES_DIR' | |
}, | |
} |