MIDCN: A multiple instance deep convolutional network for image classification

Kelei He, Jing Huo, Yinghuan Shi, Yang Gao, Dinggang Shen

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

For the image classification task, usually, the image collected in the wild contains multiple objects instead of a single dominant one. Besides, the image label is not explicitly associated with the object region, i.e., it is weakly annotated. In this paper, we propose a novel deep convolutional network for image classification under a weakly supervised condition. The proposed method, namely MIDCN, formulate the problem into Multiple Instance Learning (MIL), where each image is a bag which contains multiple instances (objects). Different with previous deep MIL methods which predict the label of each bag (i.e., image) by simply performing pooling/voting strategy over their instance (i.e., region) predictions, MIDCN directly predicts the label of a bag via bag features learned by measuring the similarities between instance features and a set of learned informative prototypes. Specifically, the prototypes are obtained by a newly proposed Global Contrast Pooling (GCP) layer which leverages instances not only coming from the current bag but also the other bags. Thus the learned bag features also contain global information of all the training bags, which is more robust and noise free. We did extensive experiments on two real-world image datasets, including both natural image dataset (PASCAL VOC 07) and pathological lung cancer image dataset, and show the results of the proposed MIDCN consistently outperforms the state-of-the-art methods.

Original languageEnglish
Title of host publicationPRICAI 2019
Subtitle of host publicationTrends in Artificial Intelligence - 16th Pacific Rim International Conference on Artificial Intelligence, Proceedings
EditorsAbhaya C. Nayak, Alok Sharma
PublisherSpringer Verlag
Pages230-243
Number of pages14
ISBN (Print)9783030299071
DOIs
Publication statusPublished - 2019 Jan 1
Externally publishedYes
Event16th Pacific Rim International Conference on Artificial Intelligence, PRICAI 2019 - Yanuka Island, Fiji
Duration: 2019 Aug 262019 Aug 30

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11670 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference16th Pacific Rim International Conference on Artificial Intelligence, PRICAI 2019
CountryFiji
CityYanuka Island
Period19/8/2619/8/30

Fingerprint

Image classification
Image Classification
Labels
Volatile organic compounds
Pooling
Prototype
Predict
Lung Cancer
Voting
Leverage
Experiments
Prediction
Experiment
Object

Keywords

  • Convolutional neural network
  • Image classification
  • Lung cancer
  • Multiple instance learning

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

He, K., Huo, J., Shi, Y., Gao, Y., & Shen, D. (2019). MIDCN: A multiple instance deep convolutional network for image classification. In A. C. Nayak, & A. Sharma (Eds.), PRICAI 2019: Trends in Artificial Intelligence - 16th Pacific Rim International Conference on Artificial Intelligence, Proceedings (pp. 230-243). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11670 LNAI). Springer Verlag. https://doi.org/10.1007/978-3-030-29908-8_19

MIDCN : A multiple instance deep convolutional network for image classification. / He, Kelei; Huo, Jing; Shi, Yinghuan; Gao, Yang; Shen, Dinggang.

PRICAI 2019: Trends in Artificial Intelligence - 16th Pacific Rim International Conference on Artificial Intelligence, Proceedings. ed. / Abhaya C. Nayak; Alok Sharma. Springer Verlag, 2019. p. 230-243 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11670 LNAI).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

He, K, Huo, J, Shi, Y, Gao, Y & Shen, D 2019, MIDCN: A multiple instance deep convolutional network for image classification. in AC Nayak & A Sharma (eds), PRICAI 2019: Trends in Artificial Intelligence - 16th Pacific Rim International Conference on Artificial Intelligence, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11670 LNAI, Springer Verlag, pp. 230-243, 16th Pacific Rim International Conference on Artificial Intelligence, PRICAI 2019, Yanuka Island, Fiji, 19/8/26. https://doi.org/10.1007/978-3-030-29908-8_19
He K, Huo J, Shi Y, Gao Y, Shen D. MIDCN: A multiple instance deep convolutional network for image classification. In Nayak AC, Sharma A, editors, PRICAI 2019: Trends in Artificial Intelligence - 16th Pacific Rim International Conference on Artificial Intelligence, Proceedings. Springer Verlag. 2019. p. 230-243. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-29908-8_19
He, Kelei ; Huo, Jing ; Shi, Yinghuan ; Gao, Yang ; Shen, Dinggang. / MIDCN : A multiple instance deep convolutional network for image classification. PRICAI 2019: Trends in Artificial Intelligence - 16th Pacific Rim International Conference on Artificial Intelligence, Proceedings. editor / Abhaya C. Nayak ; Alok Sharma. Springer Verlag, 2019. pp. 230-243 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{4b711365f3454254882517dc4724375f,
title = "MIDCN: A multiple instance deep convolutional network for image classification",
abstract = "For the image classification task, usually, the image collected in the wild contains multiple objects instead of a single dominant one. Besides, the image label is not explicitly associated with the object region, i.e., it is weakly annotated. In this paper, we propose a novel deep convolutional network for image classification under a weakly supervised condition. The proposed method, namely MIDCN, formulate the problem into Multiple Instance Learning (MIL), where each image is a bag which contains multiple instances (objects). Different with previous deep MIL methods which predict the label of each bag (i.e., image) by simply performing pooling/voting strategy over their instance (i.e., region) predictions, MIDCN directly predicts the label of a bag via bag features learned by measuring the similarities between instance features and a set of learned informative prototypes. Specifically, the prototypes are obtained by a newly proposed Global Contrast Pooling (GCP) layer which leverages instances not only coming from the current bag but also the other bags. Thus the learned bag features also contain global information of all the training bags, which is more robust and noise free. We did extensive experiments on two real-world image datasets, including both natural image dataset (PASCAL VOC 07) and pathological lung cancer image dataset, and show the results of the proposed MIDCN consistently outperforms the state-of-the-art methods.",
keywords = "Convolutional neural network, Image classification, Lung cancer, Multiple instance learning",
author = "Kelei He and Jing Huo and Yinghuan Shi and Yang Gao and Dinggang Shen",
year = "2019",
month = "1",
day = "1",
doi = "10.1007/978-3-030-29908-8_19",
language = "English",
isbn = "9783030299071",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "230--243",
editor = "Nayak, {Abhaya C.} and Alok Sharma",
booktitle = "PRICAI 2019",

}

TY - GEN

T1 - MIDCN

T2 - A multiple instance deep convolutional network for image classification

AU - He, Kelei

AU - Huo, Jing

AU - Shi, Yinghuan

AU - Gao, Yang

AU - Shen, Dinggang

PY - 2019/1/1

Y1 - 2019/1/1

N2 - For the image classification task, usually, the image collected in the wild contains multiple objects instead of a single dominant one. Besides, the image label is not explicitly associated with the object region, i.e., it is weakly annotated. In this paper, we propose a novel deep convolutional network for image classification under a weakly supervised condition. The proposed method, namely MIDCN, formulate the problem into Multiple Instance Learning (MIL), where each image is a bag which contains multiple instances (objects). Different with previous deep MIL methods which predict the label of each bag (i.e., image) by simply performing pooling/voting strategy over their instance (i.e., region) predictions, MIDCN directly predicts the label of a bag via bag features learned by measuring the similarities between instance features and a set of learned informative prototypes. Specifically, the prototypes are obtained by a newly proposed Global Contrast Pooling (GCP) layer which leverages instances not only coming from the current bag but also the other bags. Thus the learned bag features also contain global information of all the training bags, which is more robust and noise free. We did extensive experiments on two real-world image datasets, including both natural image dataset (PASCAL VOC 07) and pathological lung cancer image dataset, and show the results of the proposed MIDCN consistently outperforms the state-of-the-art methods.

AB - For the image classification task, usually, the image collected in the wild contains multiple objects instead of a single dominant one. Besides, the image label is not explicitly associated with the object region, i.e., it is weakly annotated. In this paper, we propose a novel deep convolutional network for image classification under a weakly supervised condition. The proposed method, namely MIDCN, formulate the problem into Multiple Instance Learning (MIL), where each image is a bag which contains multiple instances (objects). Different with previous deep MIL methods which predict the label of each bag (i.e., image) by simply performing pooling/voting strategy over their instance (i.e., region) predictions, MIDCN directly predicts the label of a bag via bag features learned by measuring the similarities between instance features and a set of learned informative prototypes. Specifically, the prototypes are obtained by a newly proposed Global Contrast Pooling (GCP) layer which leverages instances not only coming from the current bag but also the other bags. Thus the learned bag features also contain global information of all the training bags, which is more robust and noise free. We did extensive experiments on two real-world image datasets, including both natural image dataset (PASCAL VOC 07) and pathological lung cancer image dataset, and show the results of the proposed MIDCN consistently outperforms the state-of-the-art methods.

KW - Convolutional neural network

KW - Image classification

KW - Lung cancer

KW - Multiple instance learning

UR - http://www.scopus.com/inward/record.url?scp=85072869112&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85072869112&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-29908-8_19

DO - 10.1007/978-3-030-29908-8_19

M3 - Conference contribution

AN - SCOPUS:85072869112

SN - 9783030299071

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

SP - 230

EP - 243

BT - PRICAI 2019

A2 - Nayak, Abhaya C.

A2 - Sharma, Alok

PB - Springer Verlag

ER -