pytorch-semantic-segmentation icon indicating copy to clipboard operation
pytorch-semantic-segmentation copied to clipboard

Published 20 hours ago • verified publisher icon mapleneverfade

About the dataset

Open YOKE opened this issue 6 years ago • 2 comments

I'd like to implement the network on a dataset similar to cityscape, thus could you tell me how to get the avgClassSize("bicycle" : 4672.3249222261 , "caravan" : 36771.8241758242 ... exactly it's not the complete)? And can you explain the difference between id and trainid for me?

YOKE avatar Jul 20 '18 12:07 YOKE

When I try to modify my data labels in ./utils/cityscapes/helpers/labels,and NUM_CLASSES in options, error is "Failed to find all Cityscapes modules". I can't figure it out. Can you help me? Thank you very much!

YOKE avatar Jul 20 '18 12:07 YOKE

1、'avgClassSize'? Do you mean weight of each class? If it is , image 2、Each class has an unique ID,but for our experiment, not every class is required specified. For example, trainId background maybe include tree、sky、road in which is not our point to focus. I'm not sure this is a clear explanation. I suggest you to visit www.cityscapes-dataset.com to see more details about cityscape.

mapleneverfade avatar Jul 23 '18 08:07 mapleneverfade