Saved in:
Bibliographic Details
Main Authors: Bagchi, Anurag, Bao, Zhipeng, Wang, Yu-Xiong, Tokmakov, Pavel, Hebert, Martial
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2410.23287
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912524029394944
author Bagchi, Anurag
Bao, Zhipeng
Wang, Yu-Xiong
Tokmakov, Pavel
Hebert, Martial
author_facet Bagchi, Anurag
Bao, Zhipeng
Wang, Yu-Xiong
Tokmakov, Pavel
Hebert, Martial
contents We present REM, a framework for segmenting a wide range of concepts in video that can be described through natural language. Our method leverages the universal visual-language mapping learned by video diffusion models on Internet-scale data by fine-tuning them on small-scale Referring Object Segmentation datasets. Our key insight is to preserve the entirety of the generative model's architecture by shifting its objective from predicting noise to predicting mask latents. The resulting model can accurately segment rare and unseen objects, despite only being trained on a limited set of categories. Additionally, it can effortlessly generalize to non-object dynamic concepts, such as smoke or raindrops, as demonstrated in our new benchmark for Referring Video Process Segmentation (Ref-VPS). REM performs on par with the state-of-the-art on in-domain datasets, like Ref-DAVIS, while outperforming them by up to 12 IoU points out-of-domain, leveraging the power of generative pre-training. We also show that advancements in video generation directly improve segmentation.
format Preprint
id arxiv_https___arxiv_org_abs_2410_23287
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle ReferEverything: Towards Segmenting Everything We Can Speak of in Videos
Bagchi, Anurag
Bao, Zhipeng
Wang, Yu-Xiong
Tokmakov, Pavel
Hebert, Martial
Computer Vision and Pattern Recognition
We present REM, a framework for segmenting a wide range of concepts in video that can be described through natural language. Our method leverages the universal visual-language mapping learned by video diffusion models on Internet-scale data by fine-tuning them on small-scale Referring Object Segmentation datasets. Our key insight is to preserve the entirety of the generative model's architecture by shifting its objective from predicting noise to predicting mask latents. The resulting model can accurately segment rare and unseen objects, despite only being trained on a limited set of categories. Additionally, it can effortlessly generalize to non-object dynamic concepts, such as smoke or raindrops, as demonstrated in our new benchmark for Referring Video Process Segmentation (Ref-VPS). REM performs on par with the state-of-the-art on in-domain datasets, like Ref-DAVIS, while outperforming them by up to 12 IoU points out-of-domain, leveraging the power of generative pre-training. We also show that advancements in video generation directly improve segmentation.
title ReferEverything: Towards Segmenting Everything We Can Speak of in Videos
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2410.23287