Heat diffusion long-short term memory learning for 3D shape analysis

Fan Zhu, Jin Xie, Yi Fang

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

Abstract

The heat kernel is a fundamental solution in mathematical physics to distribution measurement of heat energy within a fixed region over time, and due to its unique property of being invariant to isometric transformations, the heat kernel has been an effective feature descriptor for spectral shape analysis. The majority of prior heat kernel-based strategies of building 3D shape representations fail to investigate the temporal dynamics of heat flows on 3D shape surfaces over time. In this work, we address the temporal dynamics of heat flows on 3D shapes using the long-short term memory (LSTM).We guide 3D shape descriptors toward discriminative representations by feeding heat distributions throughout time as inputs to units of heat diffusion LSTM (HD-LSTM) blocks with a supervised learning structure. We further extend HD-LSTM to a crossdomain structure (CDHD-LSTM) for learning domain-invariant representations of multi-view data. We evaluate the effectiveness of both HD-LSTM and CDHD-LSTM on 3D shape retrieval and sketch-based 3D shape retrieval tasks respectively. Experimental results on McGill dataset and SHREC 2014 dataset suggest that both methods can achieve state-of-the-art performance.

Original languageEnglish (US)
Title of host publicationComputer Vision - 14th European Conference, ECCV 2016, Proceedings
PublisherSpringer-Verlag
Pages305-321
Number of pages17
ISBN (Print)9783319464770
DOIs
StatePublished - Jan 1 2016

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9911 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Fingerprint

Heat Diffusion
Shape Analysis
Memory Term
3D shape
Heat Kernel
Heat Flow
Retrieval
Heat
Shape Representation
Shape Descriptor
Invariant
Supervised Learning
Heat transfer
Fundamental Solution
Spectral Analysis
Isometric
Descriptors
Learning
Long short-term memory
Hot Temperature

Keywords

  • 3D shape retrieval
  • Heat kernel signature
  • Long-short term memory
  • Recurrent neural network

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Zhu, F., Xie, J., & Fang, Y. (2016). Heat diffusion long-short term memory learning for 3D shape analysis. In Computer Vision - 14th European Conference, ECCV 2016, Proceedings (pp. 305-321). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9911 LNCS). Springer-Verlag. https://doi.org/10.1007/978-3-319-46478-7_19

Heat diffusion long-short term memory learning for 3D shape analysis. / Zhu, Fan; Xie, Jin; Fang, Yi.

Computer Vision - 14th European Conference, ECCV 2016, Proceedings. Springer-Verlag, 2016. p. 305-321 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9911 LNCS).

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

Zhu, F, Xie, J & Fang, Y 2016, Heat diffusion long-short term memory learning for 3D shape analysis. in Computer Vision - 14th European Conference, ECCV 2016, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9911 LNCS, Springer-Verlag, pp. 305-321. https://doi.org/10.1007/978-3-319-46478-7_19
Zhu F, Xie J, Fang Y. Heat diffusion long-short term memory learning for 3D shape analysis. In Computer Vision - 14th European Conference, ECCV 2016, Proceedings. Springer-Verlag. 2016. p. 305-321. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-46478-7_19
Zhu, Fan ; Xie, Jin ; Fang, Yi. / Heat diffusion long-short term memory learning for 3D shape analysis. Computer Vision - 14th European Conference, ECCV 2016, Proceedings. Springer-Verlag, 2016. pp. 305-321 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{ede3d91576054b578aa309711cc1e2f2,
title = "Heat diffusion long-short term memory learning for 3D shape analysis",
abstract = "The heat kernel is a fundamental solution in mathematical physics to distribution measurement of heat energy within a fixed region over time, and due to its unique property of being invariant to isometric transformations, the heat kernel has been an effective feature descriptor for spectral shape analysis. The majority of prior heat kernel-based strategies of building 3D shape representations fail to investigate the temporal dynamics of heat flows on 3D shape surfaces over time. In this work, we address the temporal dynamics of heat flows on 3D shapes using the long-short term memory (LSTM).We guide 3D shape descriptors toward discriminative representations by feeding heat distributions throughout time as inputs to units of heat diffusion LSTM (HD-LSTM) blocks with a supervised learning structure. We further extend HD-LSTM to a crossdomain structure (CDHD-LSTM) for learning domain-invariant representations of multi-view data. We evaluate the effectiveness of both HD-LSTM and CDHD-LSTM on 3D shape retrieval and sketch-based 3D shape retrieval tasks respectively. Experimental results on McGill dataset and SHREC 2014 dataset suggest that both methods can achieve state-of-the-art performance.",
keywords = "3D shape retrieval, Heat kernel signature, Long-short term memory, Recurrent neural network",
author = "Fan Zhu and Jin Xie and Yi Fang",
year = "2016",
month = "1",
day = "1",
doi = "10.1007/978-3-319-46478-7_19",
language = "English (US)",
isbn = "9783319464770",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer-Verlag",
pages = "305--321",
booktitle = "Computer Vision - 14th European Conference, ECCV 2016, Proceedings",

}

TY - GEN

T1 - Heat diffusion long-short term memory learning for 3D shape analysis

AU - Zhu, Fan

AU - Xie, Jin

AU - Fang, Yi

PY - 2016/1/1

Y1 - 2016/1/1

N2 - The heat kernel is a fundamental solution in mathematical physics to distribution measurement of heat energy within a fixed region over time, and due to its unique property of being invariant to isometric transformations, the heat kernel has been an effective feature descriptor for spectral shape analysis. The majority of prior heat kernel-based strategies of building 3D shape representations fail to investigate the temporal dynamics of heat flows on 3D shape surfaces over time. In this work, we address the temporal dynamics of heat flows on 3D shapes using the long-short term memory (LSTM).We guide 3D shape descriptors toward discriminative representations by feeding heat distributions throughout time as inputs to units of heat diffusion LSTM (HD-LSTM) blocks with a supervised learning structure. We further extend HD-LSTM to a crossdomain structure (CDHD-LSTM) for learning domain-invariant representations of multi-view data. We evaluate the effectiveness of both HD-LSTM and CDHD-LSTM on 3D shape retrieval and sketch-based 3D shape retrieval tasks respectively. Experimental results on McGill dataset and SHREC 2014 dataset suggest that both methods can achieve state-of-the-art performance.

AB - The heat kernel is a fundamental solution in mathematical physics to distribution measurement of heat energy within a fixed region over time, and due to its unique property of being invariant to isometric transformations, the heat kernel has been an effective feature descriptor for spectral shape analysis. The majority of prior heat kernel-based strategies of building 3D shape representations fail to investigate the temporal dynamics of heat flows on 3D shape surfaces over time. In this work, we address the temporal dynamics of heat flows on 3D shapes using the long-short term memory (LSTM).We guide 3D shape descriptors toward discriminative representations by feeding heat distributions throughout time as inputs to units of heat diffusion LSTM (HD-LSTM) blocks with a supervised learning structure. We further extend HD-LSTM to a crossdomain structure (CDHD-LSTM) for learning domain-invariant representations of multi-view data. We evaluate the effectiveness of both HD-LSTM and CDHD-LSTM on 3D shape retrieval and sketch-based 3D shape retrieval tasks respectively. Experimental results on McGill dataset and SHREC 2014 dataset suggest that both methods can achieve state-of-the-art performance.

KW - 3D shape retrieval

KW - Heat kernel signature

KW - Long-short term memory

KW - Recurrent neural network

UR - http://www.scopus.com/inward/record.url?scp=84990041382&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84990041382&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-46478-7_19

DO - 10.1007/978-3-319-46478-7_19

M3 - Conference contribution

SN - 9783319464770

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

SP - 305

EP - 321

BT - Computer Vision - 14th European Conference, ECCV 2016, Proceedings

PB - Springer-Verlag

ER -