Segment, Select, Correct: A Framework for Weakly-Supervised Referring Segmentation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Eiras, Francisco, Oksuz, Kemal, Bibi, Adel, Torr, Philip H. S., Dokania, Puneet K.
Natura: Preprint
Pubblicazione: 2023
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909290946625536
author Eiras, Francisco
Oksuz, Kemal
Bibi, Adel
Torr, Philip H. S.
Dokania, Puneet K.
author_facet Eiras, Francisco
Oksuz, Kemal
Bibi, Adel
Torr, Philip H. S.
Dokania, Puneet K.
contents Referring Image Segmentation (RIS) - the problem of identifying objects in images through natural language sentences - is a challenging task currently mostly solved through supervised learning. However, while collecting referred annotation masks is a time-consuming process, the few existing weakly-supervised and zero-shot approaches fall significantly short in performance compared to fully-supervised learning ones. To bridge the performance gap without mask annotations, we propose a novel weakly-supervised framework that tackles RIS by decomposing it into three steps: obtaining instance masks for the object mentioned in the referencing instruction (segment), using zero-shot learning to select a potentially correct mask for the given instruction (select), and bootstrapping a model which allows for fixing the mistakes of zero-shot selection (correct). In our experiments, using only the first two steps (zero-shot segment and select) outperforms other zero-shot baselines by as much as 16.5%, while our full method improves upon this much stronger baseline and sets the new state-of-the-art for weakly-supervised RIS, reducing the gap between the weakly-supervised and fully-supervised methods in some cases from around 33% to as little as 7%. Code is available at https://github.com/fgirbal/segment-select-correct.
format Preprint
id arxiv_https___arxiv_org_abs_2310_13479
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Segment, Select, Correct: A Framework for Weakly-Supervised Referring Segmentation
Eiras, Francisco
Oksuz, Kemal
Bibi, Adel
Torr, Philip H. S.
Dokania, Puneet K.
Computer Vision and Pattern Recognition
Machine Learning
Referring Image Segmentation (RIS) - the problem of identifying objects in images through natural language sentences - is a challenging task currently mostly solved through supervised learning. However, while collecting referred annotation masks is a time-consuming process, the few existing weakly-supervised and zero-shot approaches fall significantly short in performance compared to fully-supervised learning ones. To bridge the performance gap without mask annotations, we propose a novel weakly-supervised framework that tackles RIS by decomposing it into three steps: obtaining instance masks for the object mentioned in the referencing instruction (segment), using zero-shot learning to select a potentially correct mask for the given instruction (select), and bootstrapping a model which allows for fixing the mistakes of zero-shot selection (correct). In our experiments, using only the first two steps (zero-shot segment and select) outperforms other zero-shot baselines by as much as 16.5%, while our full method improves upon this much stronger baseline and sets the new state-of-the-art for weakly-supervised RIS, reducing the gap between the weakly-supervised and fully-supervised methods in some cases from around 33% to as little as 7%. Code is available at https://github.com/fgirbal/segment-select-correct.
title Segment, Select, Correct: A Framework for Weakly-Supervised Referring Segmentation
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2310.13479