Domain Adaptive Pre-trained Model for Mushroom Image Classification

Yifei SHEN; Zhuo LI; Yu YANG; Jiaxing SHEN

doi:10.1007/978-3-031-46674-8_20

Domain Adaptive Pre-trained Model for Mushroom Image Classification

Yifei SHEN^*, Zhuo LI, Yu YANG, Jiaxing SHEN

^*Corresponding author for this work

Department of Computing and Decision Sciences

Research output: Book Chapters | Papers in Conference Proceedings › Conference paper (refereed) › Research › peer-review

Abstract

Mushroom is highly diverse in morphology and colors, and difficult for ordinary people to discriminate between them. The lack of high-quality labeled mushroom datasets is one of the bottlenecks restricting cutting-edge image recognition models to achieve high performance in mushroom image recognition. To address the limitation, in this paper, we will introduce a large mushroom dataset, called Mushroom-23, constructed by us. It collects over 35000 labeled mushroom images belonging to 203 popular mushroom species in Hong Kong. Along with constructing the new mushroom dataset, we also propose the domain adaptive pre-trained (DAPT) model to make the state-of-the-art Vision Transformer (ViT) adaptive to specific mushroom recognition. The DAPT is first pre-trained on ImageNet and Danish Fungi (DF20) datasets, then fine-tuned on the collected Mushroom-23 dataset to gain the capability of different categories of mushrooms. Extensive experimental results show DAPT outperforms all baseline models by a large margin in terms of Accuracy and Macro F1.

Original language	English
Title of host publication	Advanced Data Mining and Applications :19th International Conference, ADMA 2023, Shenyang, China, August 21–23, 2023, Proceedings, Part IV
Editors	Xiaochun YANG, Heru SUHARTANTO, Guoren WANG, Bin WANG, Jing JIANG, Bing LI, Huaijie ZHU, Ningning CUI
Publisher	Springer
Chapter	20
Pages	284-296
Number of pages	13
ISBN (Electronic)	9783031466748
ISBN (Print)	9783031466731
DOIs	https://doi.org/10.1007/978-3-031-46674-8_20
Publication status	Published - 2023
Event	19th International Conference on Advanced Data Mining and Applications, ADMA 2023 - Shenyang, China Duration: 21 Aug 2023 → 23 Aug 2023

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	14179
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	19th International Conference on Advanced Data Mining and Applications, ADMA 2023
Country/Territory	China
City	Shenyang
Period	21/08/23 → 23/08/23

Bibliographical note

Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023.

Funding

This research was partially conducted in Research Institute for Artificial Intelligence of Things (RIAIoT) at PolyU and supported by PolyU Research and Innovation Office (No. BD4A), Hong Kong, China, and the Ph.D. Scientific Research Foundation, The Chongqing University of Posts and Telecommunications (No. E012A2022026), Chongqing, China, and benefited from the financial support of the Hong Kong Institute of Business Studies, Faculty of Business, Lingnan University, Hong Kong, China.

Keywords

Mushroom Recognition
Domain Adaption
Pre-trained Model
Vision Transformer

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1007/978-3-031-46674-8_20

Cite this

SHEN, Y., LI, Z., YANG, Y., & SHEN, J. (2023). Domain Adaptive Pre-trained Model for Mushroom Image Classification. In X. YANG, H. SUHARTANTO, G. WANG, B. WANG, J. JIANG, B. LI, H. ZHU, & N. CUI (Eds.), Advanced Data Mining and Applications :19th International Conference, ADMA 2023, Shenyang, China, August 21–23, 2023, Proceedings, Part IV (pp. 284-296). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14179). Springer. https://doi.org/10.1007/978-3-031-46674-8_20

SHEN, Yifei ; LI, Zhuo ; YANG, Yu et al. / Domain Adaptive Pre-trained Model for Mushroom Image Classification. Advanced Data Mining and Applications :19th International Conference, ADMA 2023, Shenyang, China, August 21–23, 2023, Proceedings, Part IV. editor / Xiaochun YANG ; Heru SUHARTANTO ; Guoren WANG ; Bin WANG ; Jing JIANG ; Bing LI ; Huaijie ZHU ; Ningning CUI. Springer, 2023. pp. 284-296 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{19e509dd002844079a01cab2572c1f55,

title = "Domain Adaptive Pre-trained Model for Mushroom Image Classification",

abstract = "Mushroom is highly diverse in morphology and colors, and difficult for ordinary people to discriminate between them. The lack of high-quality labeled mushroom datasets is one of the bottlenecks restricting cutting-edge image recognition models to achieve high performance in mushroom image recognition. To address the limitation, in this paper, we will introduce a large mushroom dataset, called Mushroom-23, constructed by us. It collects over 35000 labeled mushroom images belonging to 203 popular mushroom species in Hong Kong. Along with constructing the new mushroom dataset, we also propose the domain adaptive pre-trained (DAPT) model to make the state-of-the-art Vision Transformer (ViT) adaptive to specific mushroom recognition. The DAPT is first pre-trained on ImageNet and Danish Fungi (DF20) datasets, then fine-tuned on the collected Mushroom-23 dataset to gain the capability of different categories of mushrooms. Extensive experimental results show DAPT outperforms all baseline models by a large margin in terms of Accuracy and Macro F1.",

keywords = "Mushroom Recognition, Domain Adaption, Pre-trained Model, Vision Transformer",

author = "Yifei SHEN and Zhuo LI and Yu YANG and Jiaxing SHEN",

note = "Publisher Copyright: {\textcopyright} The Author(s), under exclusive license to Springer Nature Switzerland AG 2023.; 19th International Conference on Advanced Data Mining and Applications, ADMA 2023 ; Conference date: 21-08-2023 Through 23-08-2023",

year = "2023",

doi = "10.1007/978-3-031-46674-8_20",

language = "English",

isbn = "9783031466731",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer",

pages = "284--296",

editor = "Xiaochun YANG and Heru SUHARTANTO and Guoren WANG and Bin WANG and Jing JIANG and Bing LI and Huaijie ZHU and Ningning CUI",

booktitle = "Advanced Data Mining and Applications :19th International Conference, ADMA 2023, Shenyang, China, August 21–23, 2023, Proceedings, Part IV",

}

SHEN, Y, LI, Z, YANG, Y & SHEN, J 2023, Domain Adaptive Pre-trained Model for Mushroom Image Classification. in X YANG, H SUHARTANTO, G WANG, B WANG, J JIANG, B LI, H ZHU & N CUI (eds), Advanced Data Mining and Applications :19th International Conference, ADMA 2023, Shenyang, China, August 21–23, 2023, Proceedings, Part IV. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 14179, Springer, pp. 284-296, 19th International Conference on Advanced Data Mining and Applications, ADMA 2023, Shenyang, China, 21/08/23. https://doi.org/10.1007/978-3-031-46674-8_20

Domain Adaptive Pre-trained Model for Mushroom Image Classification. / SHEN, Yifei; LI, Zhuo; YANG, Yu et al.
Advanced Data Mining and Applications :19th International Conference, ADMA 2023, Shenyang, China, August 21–23, 2023, Proceedings, Part IV. ed. / Xiaochun YANG; Heru SUHARTANTO; Guoren WANG; Bin WANG; Jing JIANG; Bing LI; Huaijie ZHU; Ningning CUI. Springer, 2023. p. 284-296 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14179).

Research output: Book Chapters | Papers in Conference Proceedings › Conference paper (refereed) › Research › peer-review

TY - GEN

T1 - Domain Adaptive Pre-trained Model for Mushroom Image Classification

AU - SHEN, Yifei

AU - LI, Zhuo

AU - YANG, Yu

AU - SHEN, Jiaxing

N1 - Publisher Copyright: © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023.

PY - 2023

Y1 - 2023

N2 - Mushroom is highly diverse in morphology and colors, and difficult for ordinary people to discriminate between them. The lack of high-quality labeled mushroom datasets is one of the bottlenecks restricting cutting-edge image recognition models to achieve high performance in mushroom image recognition. To address the limitation, in this paper, we will introduce a large mushroom dataset, called Mushroom-23, constructed by us. It collects over 35000 labeled mushroom images belonging to 203 popular mushroom species in Hong Kong. Along with constructing the new mushroom dataset, we also propose the domain adaptive pre-trained (DAPT) model to make the state-of-the-art Vision Transformer (ViT) adaptive to specific mushroom recognition. The DAPT is first pre-trained on ImageNet and Danish Fungi (DF20) datasets, then fine-tuned on the collected Mushroom-23 dataset to gain the capability of different categories of mushrooms. Extensive experimental results show DAPT outperforms all baseline models by a large margin in terms of Accuracy and Macro F1.

AB - Mushroom is highly diverse in morphology and colors, and difficult for ordinary people to discriminate between them. The lack of high-quality labeled mushroom datasets is one of the bottlenecks restricting cutting-edge image recognition models to achieve high performance in mushroom image recognition. To address the limitation, in this paper, we will introduce a large mushroom dataset, called Mushroom-23, constructed by us. It collects over 35000 labeled mushroom images belonging to 203 popular mushroom species in Hong Kong. Along with constructing the new mushroom dataset, we also propose the domain adaptive pre-trained (DAPT) model to make the state-of-the-art Vision Transformer (ViT) adaptive to specific mushroom recognition. The DAPT is first pre-trained on ImageNet and Danish Fungi (DF20) datasets, then fine-tuned on the collected Mushroom-23 dataset to gain the capability of different categories of mushrooms. Extensive experimental results show DAPT outperforms all baseline models by a large margin in terms of Accuracy and Macro F1.

KW - Mushroom Recognition

KW - Domain Adaption

KW - Pre-trained Model

KW - Vision Transformer

UR - http://www.scopus.com/inward/record.url?scp=85177470517&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-46674-8_20

DO - 10.1007/978-3-031-46674-8_20

M3 - Conference paper (refereed)

SN - 9783031466731

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 284

EP - 296

BT - Advanced Data Mining and Applications :19th International Conference, ADMA 2023, Shenyang, China, August 21–23, 2023, Proceedings, Part IV

A2 - YANG, Xiaochun

A2 - SUHARTANTO, Heru

A2 - WANG, Guoren

A2 - WANG, Bin

A2 - JIANG, Jing

A2 - LI, Bing

A2 - ZHU, Huaijie

A2 - CUI, Ningning

PB - Springer

T2 - 19th International Conference on Advanced Data Mining and Applications, ADMA 2023

Y2 - 21 August 2023 through 23 August 2023

ER -

SHEN Y, LI Z, YANG Y, SHEN J. Domain Adaptive Pre-trained Model for Mushroom Image Classification. In YANG X, SUHARTANTO H, WANG G, WANG B, JIANG J, LI B, ZHU H, CUI N, editors, Advanced Data Mining and Applications :19th International Conference, ADMA 2023, Shenyang, China, August 21–23, 2023, Proceedings, Part IV. Springer. 2023. p. 284-296. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). Epub 2023 Nov 5. doi: 10.1007/978-3-031-46674-8_20