Has anyone successfully trained the model from
here? Could you tell me what you did/used to produce the databases for the data and labels? I'm having a difficult time trying to figure out what I should be using as the training label (I'm using the PASCAL-Context dataset). I've been using a set of the original images as the data and a set from the semantic class segmentation set (produced from the other images), but all I've gotten is errors.