Pre‐trained SAM as data augmentation for image segmentation

Abstract Data augmentation plays an important role in training deep neural model by expanding the size and diversity of the dataset. Initially, data augmentation mainly involved some simple transformations of images. Later, in order to increase the diversity and complexity of data, more advanced met...

Full description

Saved in:
Bibliographic Details
Main Authors: Junjun Wu, Yunbo Rao, Shaoning Zeng, Bob Zhang
Format: Article
Language:English
Published: Wiley 2025-02-01
Series:CAAI Transactions on Intelligence Technology
Subjects:
Online Access:https://doi.org/10.1049/cit2.12381
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849324008222228480
author Junjun Wu
Yunbo Rao
Shaoning Zeng
Bob Zhang
author_facet Junjun Wu
Yunbo Rao
Shaoning Zeng
Bob Zhang
author_sort Junjun Wu
collection DOAJ
description Abstract Data augmentation plays an important role in training deep neural model by expanding the size and diversity of the dataset. Initially, data augmentation mainly involved some simple transformations of images. Later, in order to increase the diversity and complexity of data, more advanced methods appeared and evolved to sophisticated generative models. However, these methods required a mass of computation of training or searching. In this paper, a novel training‐free method that utilises the Pre‐Trained Segment Anything Model (SAM) model as a data augmentation tool (PTSAM‐DA) is proposed to generate the augmented annotations for images. Without the need for training, it obtains prompt boxes from the original annotations and then feeds the boxes to the pre‐trained SAM to generate diverse and improved annotations. In this way, annotations are augmented more ingenious than simple manipulations without incurring huge computation for training a data augmentation model. Multiple comparative experiments on three datasets are conducted, including an in‐house dataset, ADE20K and COCO2017. On this in‐house dataset, namely Agricultural Plot Segmentation Dataset, maximum improvements of 3.77% and 8.92% are gained in two mainstream metrics, mIoU and mAcc, respectively. Consequently, large vision models like SAM are proven to be promising not only in image segmentation but also in data augmentation.
format Article
id doaj-art-cd82b1dbf933483ea7a85828b01e0669
institution Kabale University
issn 2468-2322
language English
publishDate 2025-02-01
publisher Wiley
record_format Article
series CAAI Transactions on Intelligence Technology
spelling doaj-art-cd82b1dbf933483ea7a85828b01e06692025-08-20T03:48:51ZengWileyCAAI Transactions on Intelligence Technology2468-23222025-02-0110126828210.1049/cit2.12381Pre‐trained SAM as data augmentation for image segmentationJunjun Wu0Yunbo Rao1Shaoning Zeng2Bob Zhang3Yangtze Delta Region Institute (Huzhou) University of Electronic Science and Technology of China Huzhou ChinaSchool of Information and Software Engineering University of Electronic Science and Technology of China Chengdu ChinaYangtze Delta Region Institute (Huzhou) University of Electronic Science and Technology of China Huzhou ChinaPattern Analysis and Machine Intelligence Research Group Department of Computer and Information Science University of Macau Macau ChinaAbstract Data augmentation plays an important role in training deep neural model by expanding the size and diversity of the dataset. Initially, data augmentation mainly involved some simple transformations of images. Later, in order to increase the diversity and complexity of data, more advanced methods appeared and evolved to sophisticated generative models. However, these methods required a mass of computation of training or searching. In this paper, a novel training‐free method that utilises the Pre‐Trained Segment Anything Model (SAM) model as a data augmentation tool (PTSAM‐DA) is proposed to generate the augmented annotations for images. Without the need for training, it obtains prompt boxes from the original annotations and then feeds the boxes to the pre‐trained SAM to generate diverse and improved annotations. In this way, annotations are augmented more ingenious than simple manipulations without incurring huge computation for training a data augmentation model. Multiple comparative experiments on three datasets are conducted, including an in‐house dataset, ADE20K and COCO2017. On this in‐house dataset, namely Agricultural Plot Segmentation Dataset, maximum improvements of 3.77% and 8.92% are gained in two mainstream metrics, mIoU and mAcc, respectively. Consequently, large vision models like SAM are proven to be promising not only in image segmentation but also in data augmentation.https://doi.org/10.1049/cit2.12381data augmentationimage segmentationlarge modelsegment anything model
spellingShingle Junjun Wu
Yunbo Rao
Shaoning Zeng
Bob Zhang
Pre‐trained SAM as data augmentation for image segmentation
CAAI Transactions on Intelligence Technology
data augmentation
image segmentation
large model
segment anything model
title Pre‐trained SAM as data augmentation for image segmentation
title_full Pre‐trained SAM as data augmentation for image segmentation
title_fullStr Pre‐trained SAM as data augmentation for image segmentation
title_full_unstemmed Pre‐trained SAM as data augmentation for image segmentation
title_short Pre‐trained SAM as data augmentation for image segmentation
title_sort pre trained sam as data augmentation for image segmentation
topic data augmentation
image segmentation
large model
segment anything model
url https://doi.org/10.1049/cit2.12381
work_keys_str_mv AT junjunwu pretrainedsamasdataaugmentationforimagesegmentation
AT yunborao pretrainedsamasdataaugmentationforimagesegmentation
AT shaoningzeng pretrainedsamasdataaugmentationforimagesegmentation
AT bobzhang pretrainedsamasdataaugmentationforimagesegmentation