Superpixel Positional Encoding to Improve ViT-based Semantic Segmentation Models.

Roberto Amoroso Matteo Tomei Lorenzo Baraldi Rita Cucchiara

Published in: BMVC (2023)

Keyphrases

semantic segmentation
superpixels
probabilistic model
conditional random fields
viewpoint
input image
street scenes
prior knowledge
image features
object categories
weakly supervised