Salvato in:
Dettagli Bibliografici
Autori principali: Hwang, Dongjun, Kim, Yejin, Lee, Minyoung, Oh, Seong Joon, Choe, Junsuk
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2410.11536
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866917005626441728
author Hwang, Dongjun
Kim, Yejin
Lee, Minyoung
Oh, Seong Joon
Choe, Junsuk
author_facet Hwang, Dongjun
Kim, Yejin
Lee, Minyoung
Oh, Seong Joon
Choe, Junsuk
contents Open-Vocabulary Segmentation (OVS) aims to segment classes that are not present in the training dataset. However, most existing studies assume that the training data is fixed in advance, overlooking more practical scenarios where new datasets are continuously collected over time. To address this, we first analyze how existing OVS models perform under such conditions. In this context, we explore several approaches such as retraining, fine-tuning, and continual learning but find that each of them has clear limitations. To address these issues, we propose ConOVS, a novel continual learning method based on a Mixture-of-Experts framework. ConOVS dynamically combines expert decoders based on the probability that an input sample belongs to the distribution of each incremental dataset. Through extensive experiments, we show that ConOVS consistently outperforms existing methods across pre-training, incremental, and zero-shot test datasets, effectively expanding the recognition capabilities of OVS models when data is collected sequentially.
format Preprint
id arxiv_https___arxiv_org_abs_2410_11536
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle OVS Meets Continual Learning: Towards Sustainable Open-Vocabulary Segmentation
Hwang, Dongjun
Kim, Yejin
Lee, Minyoung
Oh, Seong Joon
Choe, Junsuk
Computer Vision and Pattern Recognition
Open-Vocabulary Segmentation (OVS) aims to segment classes that are not present in the training dataset. However, most existing studies assume that the training data is fixed in advance, overlooking more practical scenarios where new datasets are continuously collected over time. To address this, we first analyze how existing OVS models perform under such conditions. In this context, we explore several approaches such as retraining, fine-tuning, and continual learning but find that each of them has clear limitations. To address these issues, we propose ConOVS, a novel continual learning method based on a Mixture-of-Experts framework. ConOVS dynamically combines expert decoders based on the probability that an input sample belongs to the distribution of each incremental dataset. Through extensive experiments, we show that ConOVS consistently outperforms existing methods across pre-training, incremental, and zero-shot test datasets, effectively expanding the recognition capabilities of OVS models when data is collected sequentially.
title OVS Meets Continual Learning: Towards Sustainable Open-Vocabulary Segmentation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2410.11536