A Simple Framework for Open-Vocabulary Segmentation and Detection. Jointly learns from different segmentation and detection datasets with a pre-trained text encoder for a common semantic space. Proposes decoupled decoding for foreground/background and conditioned mask decoding for box-to-mask.

Outputs 2

OpenSeeD

library

Official open-vocabulary segmentation and detection framework with interactive segmentation capabilities.

GitHub Repository

A Simple Framework for Open-Vocabulary Segmentation and Detection

paper

Proposes decoupled and conditioned decoding to unify segmentation and detection in an open-vocabulary setting.

arXiv: 2303.08131

Venue: ICCV 2023

visionopen-vocabularyopen-source