what does self.label_map[i] = -1 mean in custom.py?

[RoyZhDec]

In custom.py, if I use custom_classes(classes is not None), the code below will process:
for i, c in enumerate(self.CLASSES): if c not in class_names: self.label_map[i] = -1 else: self.label_map[i] = class_names.index(c)
For example, if I use cityscapes for training and I only want to use "road" class, all other classes I treat as background, and I hope to get a binary segmentation model. So I add one parameter "classes=['road']" to build my own cityscapes dataset as follow:
super(CityscapesDataset_2class, self).__init__( img_suffix=img_suffix, seg_map_suffix=seg_map_suffix, classes=['road'], **kwargs)
Then besides 'road' train_id will still be 0, all other classes train_id will be -1, what exactly does this mean? -1 means ignore? Or just another class? Can I get a binary model? Looking forward your reply! BIG THANKS!!

[RockeyCoss]

Sorry for late reply.
Well, we use uint8 as the data type of the annotations. Therefore, when loading the annotation, -1 is actually 255. 255 is the default ignore index, which meas classes other than road will be ignored