Remove Your Limits: This AI Approach Uses Diffusion Models to Enable Open-Vocabulary Object Segmentation

One of the main problems in object segmentation models is their limited dictionaries. There have been multiple attempts to tackle this “closed” vocabulary of object segmentation models. How can we make object segmentation models’ vocabulary an open one? Source: https://arxiv.org/abs/2303.04803ODISE is the first work to explore large-scale text-to-image diffusion models for open-vocabulary segmentation tasks. It proposes a novel pipeline to effectively leverage both text-image diffusion and discriminative models to perform open-vocabulary panoptic segmentation.