perform bad on unseen classes

cvlab-kaist / CAT-Seg

Official Implementation of "CAT-Seg🐱: Cost Aggregation for Open-Vocabulary Semantic Segmentation"

https://ku-cvlab.github.io/CAT-Seg/

MIT License

267 stars 25 forks source link

perform bad on unseen classes #14

Closed bingnoi closed 7 months ago

bingnoi commented 1 year ago

i try to train on my own datasets, it shows not bad on seen classes, but got bad results on unseen classes, mostly got 0 precision. what might be the reason?

hsshin98 commented 7 months ago

Hi, can you specify what kind of dataset or classes you are dealing with? Also, we released our updated version of CAT-Seg with lots of improvements, so you might try our updated version as well!