Margin Constraint for Low-Shot Learning

Xiaotian Wu; Yizhuo Wang

doi:10.1007/978-3-030-41299-9_1

Margin Constraint for Low-Shot Learning

Xiaotian Wu^*, Yizhuo Wang

^*Corresponding author for this work

School of Computer Science and Technology

Beijing Institute of Technology

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Low-shot learning aims to recognize novel visual categories with limited examples, which is mimicking the human visual system and remains a challenging research problem. In this paper, we introduce the margin constraint in loss function for the low-shot learning field to enhance the model’s discriminative power. Additionally, we adopt the novel categories’ normalized feature vectors as the corresponding classification weight vectors directly, in order to provide an instant classification performance on the novel categories without retraining. Experiments show that our method provides a better generalization and outperforms the previous methods on the low-shot leaning benchmarks.

Original language	English
Title of host publication	Pattern Recognition - 5th Asian Conference, ACPR 2019, Revised Selected Papers
Editors	Shivakumara Palaiahnakote, Gabriella Sanniti di Baja, Liang Wang, Wei Qi Yan
Publisher	Springer
Pages	3-14
Number of pages	12
ISBN (Print)	9783030412982
DOIs	https://doi.org/10.1007/978-3-030-41299-9_1
Publication status	Published - 2020
Event	5th Asian Conference on Pattern Recognition, ACPR 2019 - Auckland, New Zealand Duration: 26 Nov 2019 → 29 Nov 2019

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	12047 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	5th Asian Conference on Pattern Recognition, ACPR 2019
Country/Territory	New Zealand
City	Auckland
Period	26/11/19 → 29/11/19

Keywords

Low-shot learning
Margin constraint
Normalized vectors

Access to Document

10.1007/978-3-030-41299-9_1

Cite this

Wu, X., & Wang, Y. (2020). Margin Constraint for Low-Shot Learning. In S. Palaiahnakote, G. Sanniti di Baja, L. Wang, & W. Q. Yan (Eds.), Pattern Recognition - 5th Asian Conference, ACPR 2019, Revised Selected Papers (pp. 3-14). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12047 LNCS). Springer. https://doi.org/10.1007/978-3-030-41299-9_1

Wu, Xiaotian ; Wang, Yizhuo. / Margin Constraint for Low-Shot Learning. Pattern Recognition - 5th Asian Conference, ACPR 2019, Revised Selected Papers. editor / Shivakumara Palaiahnakote ; Gabriella Sanniti di Baja ; Liang Wang ; Wei Qi Yan. Springer, 2020. pp. 3-14 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{6529f5e64d66474bb0aa0837d4683679,

title = "Margin Constraint for Low-Shot Learning",

abstract = "Low-shot learning aims to recognize novel visual categories with limited examples, which is mimicking the human visual system and remains a challenging research problem. In this paper, we introduce the margin constraint in loss function for the low-shot learning field to enhance the model{\textquoteright}s discriminative power. Additionally, we adopt the novel categories{\textquoteright} normalized feature vectors as the corresponding classification weight vectors directly, in order to provide an instant classification performance on the novel categories without retraining. Experiments show that our method provides a better generalization and outperforms the previous methods on the low-shot leaning benchmarks.",

keywords = "Low-shot learning, Margin constraint, Normalized vectors",

author = "Xiaotian Wu and Yizhuo Wang",

note = "Publisher Copyright: {\textcopyright} 2020, Springer Nature Switzerland AG.; 5th Asian Conference on Pattern Recognition, ACPR 2019 ; Conference date: 26-11-2019 Through 29-11-2019",

year = "2020",

doi = "10.1007/978-3-030-41299-9_1",

language = "English",

isbn = "9783030412982",

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

publisher = "Springer",

pages = "3--14",

editor = "Shivakumara Palaiahnakote and {Sanniti di Baja}, Gabriella and Liang Wang and Yan, {Wei Qi}",

booktitle = "Pattern Recognition - 5th Asian Conference, ACPR 2019, Revised Selected Papers",

address = "Germany",

}

Wu, X & Wang, Y 2020, Margin Constraint for Low-Shot Learning. in S Palaiahnakote, G Sanniti di Baja, L Wang & WQ Yan (eds), Pattern Recognition - 5th Asian Conference, ACPR 2019, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12047 LNCS, Springer, pp. 3-14, 5th Asian Conference on Pattern Recognition, ACPR 2019, Auckland, New Zealand, 26/11/19. https://doi.org/10.1007/978-3-030-41299-9_1

Margin Constraint for Low-Shot Learning. / Wu, Xiaotian; Wang, Yizhuo.
Pattern Recognition - 5th Asian Conference, ACPR 2019, Revised Selected Papers. ed. / Shivakumara Palaiahnakote; Gabriella Sanniti di Baja; Liang Wang; Wei Qi Yan. Springer, 2020. p. 3-14 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12047 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Margin Constraint for Low-Shot Learning

AU - Wu, Xiaotian

AU - Wang, Yizhuo

PY - 2020

Y1 - 2020

N2 - Low-shot learning aims to recognize novel visual categories with limited examples, which is mimicking the human visual system and remains a challenging research problem. In this paper, we introduce the margin constraint in loss function for the low-shot learning field to enhance the model’s discriminative power. Additionally, we adopt the novel categories’ normalized feature vectors as the corresponding classification weight vectors directly, in order to provide an instant classification performance on the novel categories without retraining. Experiments show that our method provides a better generalization and outperforms the previous methods on the low-shot leaning benchmarks.

AB - Low-shot learning aims to recognize novel visual categories with limited examples, which is mimicking the human visual system and remains a challenging research problem. In this paper, we introduce the margin constraint in loss function for the low-shot learning field to enhance the model’s discriminative power. Additionally, we adopt the novel categories’ normalized feature vectors as the corresponding classification weight vectors directly, in order to provide an instant classification performance on the novel categories without retraining. Experiments show that our method provides a better generalization and outperforms the previous methods on the low-shot leaning benchmarks.

KW - Low-shot learning

KW - Margin constraint

KW - Normalized vectors

UR - http://www.scopus.com/inward/record.url?scp=85081542801&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-41299-9_1

DO - 10.1007/978-3-030-41299-9_1

M3 - Conference contribution

AN - SCOPUS:85081542801

SN - 9783030412982

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

SP - 3

EP - 14

BT - Pattern Recognition - 5th Asian Conference, ACPR 2019, Revised Selected Papers

A2 - Palaiahnakote, Shivakumara

A2 - Sanniti di Baja, Gabriella

A2 - Wang, Liang

A2 - Yan, Wei Qi

PB - Springer

T2 - 5th Asian Conference on Pattern Recognition, ACPR 2019

Y2 - 26 November 2019 through 29 November 2019

ER -

Wu X, Wang Y. Margin Constraint for Low-Shot Learning. In Palaiahnakote S, Sanniti di Baja G, Wang L, Yan WQ, editors, Pattern Recognition - 5th Asian Conference, ACPR 2019, Revised Selected Papers. Springer. 2020. p. 3-14. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-41299-9_1

Margin Constraint for Low-Shot Learning

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this