You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
You can train cityscapes images by modifying the #classes of the model and integrating with cityscapes dataloader someone's provided, and moreover, you can initialize the model with the init.caffemodel because of skipping the last layer which is dataset dependent. But I'm not sure about the validity with the COCO pre-trained init.caffemodel. According to the general cityscapes manner, I think that an imagenet pre-trained model is used to initialize.
P.S. sorry totally changed from the first comment.
Hi, @kazuto1011 Thanks for your work. I would like to know how to use your code on CityScapes? The init.caffemodel can be used for cityscapes or not?
The text was updated successfully, but these errors were encountered: