Flexible visual prompts for in-context learning in computer vision

TL;DR

This paper introduces a flexible visual prompt approach for in-context learning in image segmentation, adapting Video Object Segmentation techniques to improve performance on diverse datasets and unseen classes without extra training.

Contribution

It presents a novel adaptation of VOS methods for visual in-context learning and introduces a support set selection technique that enhances performance without additional training.

Findings

Outperforms existing segmentation methods across datasets

Excels with classes not seen during training

Support set selection improves results without extra training

Abstract

In this work, we address in-context learning (ICL) for the task of image segmentation, introducing a novel approach that adapts a modern Video Object Segmentation (VOS) technique for visual in-context learning. This adaptation is inspired by the VOS method's ability to efficiently and flexibly learn objects from a few examples. Through evaluations across a range of support set sizes and on diverse segmentation datasets, our method consistently surpasses existing techniques. Notably, it excels with data containing classes not encountered during training. Additionally, we propose a technique for support set selection, which involves choosing the most relevant images to include in this set. By employing support set selection, the performance increases for all tested methods without the need for additional training or prompt tuning. The code can be found at https://github.com/v7labs/XMem_ICL/.