Information-Theoretic Odometry Learning View Full Text


Ontology type: schema:ScholarlyArticle      Open Access: True


Article Info

DATE

2022-08-12

AUTHORS

Sen Zhang, Jing Zhang, Dacheng Tao

ABSTRACT

In this paper, we propose a unified information theoretic framework for learning-motivated methods aimed at odometry estimation, a crucial component of many robotics and vision tasks such as navigation and virtual reality where relative camera poses are required in real time. We formulate this problem as optimizing a variational information bottleneck objective function, which eliminates pose-irrelevant information from the latent representation. The proposed framework provides an elegant tool for performance evaluation and understanding in information-theoretic language. Specifically, we bound the generalization errors of the deep information bottleneck framework and the predictability of the latent representation. These provide not only a performance guarantee but also practical guidance for model design, sample collection, and sensor selection. Furthermore, the stochastic latent representation provides a natural uncertainty measure without the needs for extra structures or computations. Experiments on two well-known odometry datasets demonstrate the effectiveness of our method. More... »

PAGES

2553-2570

References to SciGraph publications

  • 2012-11-13. Visual simultaneous localization and mapping: a survey in ARTIFICIAL INTELLIGENCE REVIEW
  • 2017-06-02. Visual SLAM algorithms: a survey from 2010 to 2016 in IPSJ TRANSACTIONS ON COMPUTER VISION AND APPLICATIONS
  • 2020-11-03. RAFT: Recurrent All-Pairs Field Transforms for Optical Flow in COMPUTER VISION – ECCV 2020
  • Identifiers

    URI

    http://scigraph.springernature.com/pub.10.1007/s11263-022-01659-9

    DOI

    http://dx.doi.org/10.1007/s11263-022-01659-9

    DIMENSIONS

    https://app.dimensions.ai/details/publication/pub.1150189028


    Indexing Status Check whether this publication has been indexed by Scopus and Web Of Science using the SN Indexing Status Tool
    Incoming Citations Browse incoming citations for this publication using opencitations.net

    JSON-LD is the canonical representation for SciGraph data.

    TIP: You can open this SciGraph record using an external JSON-LD service: JSON-LD Playground Google SDTT

    [
      {
        "@context": "https://springernature.github.io/scigraph/jsonld/sgcontext.json", 
        "about": [
          {
            "id": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/08", 
            "inDefinedTermSet": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/", 
            "name": "Information and Computing Sciences", 
            "type": "DefinedTerm"
          }, 
          {
            "id": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/0801", 
            "inDefinedTermSet": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/", 
            "name": "Artificial Intelligence and Image Processing", 
            "type": "DefinedTerm"
          }
        ], 
        "author": [
          {
            "affiliation": {
              "alternateName": "The University of Sydney, Sydney, NSW, Australia", 
              "id": "http://www.grid.ac/institutes/grid.1013.3", 
              "name": [
                "The University of Sydney, Sydney, NSW, Australia"
              ], 
              "type": "Organization"
            }, 
            "familyName": "Zhang", 
            "givenName": "Sen", 
            "id": "sg:person.014105751612.08", 
            "sameAs": [
              "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.014105751612.08"
            ], 
            "type": "Person"
          }, 
          {
            "affiliation": {
              "alternateName": "The University of Sydney, Sydney, NSW, Australia", 
              "id": "http://www.grid.ac/institutes/grid.1013.3", 
              "name": [
                "The University of Sydney, Sydney, NSW, Australia"
              ], 
              "type": "Organization"
            }, 
            "familyName": "Zhang", 
            "givenName": "Jing", 
            "id": "sg:person.010767212714.33", 
            "sameAs": [
              "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.010767212714.33"
            ], 
            "type": "Person"
          }, 
          {
            "affiliation": {
              "alternateName": "The University of Sydney, Sydney, NSW, Australia", 
              "id": "http://www.grid.ac/institutes/grid.1013.3", 
              "name": [
                "The University of Sydney, Sydney, NSW, Australia"
              ], 
              "type": "Organization"
            }, 
            "familyName": "Tao", 
            "givenName": "Dacheng", 
            "id": "sg:person.0677067025.10", 
            "sameAs": [
              "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.0677067025.10"
            ], 
            "type": "Person"
          }
        ], 
        "citation": [
          {
            "id": "sg:pub.10.1007/s10462-012-9365-8", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1050263800", 
              "https://doi.org/10.1007/s10462-012-9365-8"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1186/s41074-017-0027-2", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1085758120", 
              "https://doi.org/10.1186/s41074-017-0027-2"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1007/978-3-030-58536-5_24", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1132270164", 
              "https://doi.org/10.1007/978-3-030-58536-5_24"
            ], 
            "type": "CreativeWork"
          }
        ], 
        "datePublished": "2022-08-12", 
        "datePublishedReg": "2022-08-12", 
        "description": "In this paper, we propose a unified information theoretic framework for learning-motivated methods aimed at odometry estimation, a crucial component of many robotics and vision tasks such as navigation and virtual reality where relative camera poses are required in real time. We formulate this problem as optimizing a variational information bottleneck objective function, which eliminates pose-irrelevant information from the latent representation. The proposed framework provides an elegant tool for performance evaluation and understanding in information-theoretic language. Specifically, we bound the generalization errors of the deep information bottleneck framework and the predictability of the latent representation. These provide not only a performance guarantee but also practical guidance for model design, sample collection, and sensor selection. Furthermore, the stochastic latent representation provides a natural uncertainty measure without the needs for extra structures or computations. Experiments on two well-known odometry datasets demonstrate the effectiveness of our method.", 
        "genre": "article", 
        "id": "sg:pub.10.1007/s11263-022-01659-9", 
        "isAccessibleForFree": true, 
        "isPartOf": [
          {
            "id": "sg:journal.1032807", 
            "issn": [
              "0920-5691", 
              "1573-1405"
            ], 
            "name": "International Journal of Computer Vision", 
            "publisher": "Springer Nature", 
            "type": "Periodical"
          }, 
          {
            "issueNumber": "11", 
            "type": "PublicationIssue"
          }, 
          {
            "type": "PublicationVolume", 
            "volumeNumber": "130"
          }
        ], 
        "keywords": [
          "latent representations", 
          "relative camera pose", 
          "information bottleneck framework", 
          "information-theoretic framework", 
          "vision tasks", 
          "camera pose", 
          "odometry estimation", 
          "virtual reality", 
          "performance guarantees", 
          "generalization error", 
          "sensor selection", 
          "real time", 
          "theoretic framework", 
          "model design", 
          "performance evaluation", 
          "bottleneck objective function", 
          "uncertainty measures", 
          "objective function", 
          "framework", 
          "representation", 
          "robotics", 
          "pose", 
          "navigation", 
          "dataset", 
          "guarantees", 
          "crucial component", 
          "elegant tool", 
          "task", 
          "computation", 
          "learning", 
          "language", 
          "information", 
          "tool", 
          "reality", 
          "method", 
          "collection", 
          "effectiveness", 
          "error", 
          "practical guidance", 
          "design", 
          "estimation", 
          "selection", 
          "extra structure", 
          "need", 
          "experiments", 
          "evaluation", 
          "guidance", 
          "time", 
          "predictability", 
          "components", 
          "measures", 
          "function", 
          "structure", 
          "understanding", 
          "sample collection", 
          "problem", 
          "paper"
        ], 
        "name": "Information-Theoretic Odometry Learning", 
        "pagination": "2553-2570", 
        "productId": [
          {
            "name": "dimensions_id", 
            "type": "PropertyValue", 
            "value": [
              "pub.1150189028"
            ]
          }, 
          {
            "name": "doi", 
            "type": "PropertyValue", 
            "value": [
              "10.1007/s11263-022-01659-9"
            ]
          }
        ], 
        "sameAs": [
          "https://doi.org/10.1007/s11263-022-01659-9", 
          "https://app.dimensions.ai/details/publication/pub.1150189028"
        ], 
        "sdDataset": "articles", 
        "sdDatePublished": "2022-12-01T06:45", 
        "sdLicense": "https://scigraph.springernature.com/explorer/license/", 
        "sdPublisher": {
          "name": "Springer Nature - SN SciGraph project", 
          "type": "Organization"
        }, 
        "sdSource": "s3://com-springernature-scigraph/baseset/20221201/entities/gbq_results/article/article_950.jsonl", 
        "type": "ScholarlyArticle", 
        "url": "https://doi.org/10.1007/s11263-022-01659-9"
      }
    ]
     

    Download the RDF metadata as:  json-ld nt turtle xml License info

    HOW TO GET THIS DATA PROGRAMMATICALLY:

    JSON-LD is a popular format for linked data which is fully compatible with JSON.

    curl -H 'Accept: application/ld+json' 'https://scigraph.springernature.com/pub.10.1007/s11263-022-01659-9'

    N-Triples is a line-based linked data format ideal for batch operations.

    curl -H 'Accept: application/n-triples' 'https://scigraph.springernature.com/pub.10.1007/s11263-022-01659-9'

    Turtle is a human-readable linked data format.

    curl -H 'Accept: text/turtle' 'https://scigraph.springernature.com/pub.10.1007/s11263-022-01659-9'

    RDF/XML is a standard XML format for linked data.

    curl -H 'Accept: application/rdf+xml' 'https://scigraph.springernature.com/pub.10.1007/s11263-022-01659-9'


     

    This table displays all metadata directly associated to this object as RDF triples.

    140 TRIPLES      21 PREDICATES      84 URIs      73 LITERALS      6 BLANK NODES

    Subject Predicate Object
    1 sg:pub.10.1007/s11263-022-01659-9 schema:about anzsrc-for:08
    2 anzsrc-for:0801
    3 schema:author Neeed0a17f9744ad78e6b6089518eb6c2
    4 schema:citation sg:pub.10.1007/978-3-030-58536-5_24
    5 sg:pub.10.1007/s10462-012-9365-8
    6 sg:pub.10.1186/s41074-017-0027-2
    7 schema:datePublished 2022-08-12
    8 schema:datePublishedReg 2022-08-12
    9 schema:description In this paper, we propose a unified information theoretic framework for learning-motivated methods aimed at odometry estimation, a crucial component of many robotics and vision tasks such as navigation and virtual reality where relative camera poses are required in real time. We formulate this problem as optimizing a variational information bottleneck objective function, which eliminates pose-irrelevant information from the latent representation. The proposed framework provides an elegant tool for performance evaluation and understanding in information-theoretic language. Specifically, we bound the generalization errors of the deep information bottleneck framework and the predictability of the latent representation. These provide not only a performance guarantee but also practical guidance for model design, sample collection, and sensor selection. Furthermore, the stochastic latent representation provides a natural uncertainty measure without the needs for extra structures or computations. Experiments on two well-known odometry datasets demonstrate the effectiveness of our method.
    10 schema:genre article
    11 schema:isAccessibleForFree true
    12 schema:isPartOf N74ffa4be488e46d2aa7866560f8e288c
    13 Nadf33478e02946ceabe88a5a67dc4092
    14 sg:journal.1032807
    15 schema:keywords bottleneck objective function
    16 camera pose
    17 collection
    18 components
    19 computation
    20 crucial component
    21 dataset
    22 design
    23 effectiveness
    24 elegant tool
    25 error
    26 estimation
    27 evaluation
    28 experiments
    29 extra structure
    30 framework
    31 function
    32 generalization error
    33 guarantees
    34 guidance
    35 information
    36 information bottleneck framework
    37 information-theoretic framework
    38 language
    39 latent representations
    40 learning
    41 measures
    42 method
    43 model design
    44 navigation
    45 need
    46 objective function
    47 odometry estimation
    48 paper
    49 performance evaluation
    50 performance guarantees
    51 pose
    52 practical guidance
    53 predictability
    54 problem
    55 real time
    56 reality
    57 relative camera pose
    58 representation
    59 robotics
    60 sample collection
    61 selection
    62 sensor selection
    63 structure
    64 task
    65 theoretic framework
    66 time
    67 tool
    68 uncertainty measures
    69 understanding
    70 virtual reality
    71 vision tasks
    72 schema:name Information-Theoretic Odometry Learning
    73 schema:pagination 2553-2570
    74 schema:productId N09bc70caff1f4ebfaf6c2463f9a9b1cc
    75 N70cea8d716ca4f26adb6abc429edfb17
    76 schema:sameAs https://app.dimensions.ai/details/publication/pub.1150189028
    77 https://doi.org/10.1007/s11263-022-01659-9
    78 schema:sdDatePublished 2022-12-01T06:45
    79 schema:sdLicense https://scigraph.springernature.com/explorer/license/
    80 schema:sdPublisher Ncbbffddfa8ae4b6b8409c894b6594167
    81 schema:url https://doi.org/10.1007/s11263-022-01659-9
    82 sgo:license sg:explorer/license/
    83 sgo:sdDataset articles
    84 rdf:type schema:ScholarlyArticle
    85 N09bc70caff1f4ebfaf6c2463f9a9b1cc schema:name dimensions_id
    86 schema:value pub.1150189028
    87 rdf:type schema:PropertyValue
    88 N124e9de582914687ae4ffe1fbbabd3e7 rdf:first sg:person.010767212714.33
    89 rdf:rest N3e06d819e7454cd9addd93049d33880d
    90 N3e06d819e7454cd9addd93049d33880d rdf:first sg:person.0677067025.10
    91 rdf:rest rdf:nil
    92 N70cea8d716ca4f26adb6abc429edfb17 schema:name doi
    93 schema:value 10.1007/s11263-022-01659-9
    94 rdf:type schema:PropertyValue
    95 N74ffa4be488e46d2aa7866560f8e288c schema:issueNumber 11
    96 rdf:type schema:PublicationIssue
    97 Nadf33478e02946ceabe88a5a67dc4092 schema:volumeNumber 130
    98 rdf:type schema:PublicationVolume
    99 Ncbbffddfa8ae4b6b8409c894b6594167 schema:name Springer Nature - SN SciGraph project
    100 rdf:type schema:Organization
    101 Neeed0a17f9744ad78e6b6089518eb6c2 rdf:first sg:person.014105751612.08
    102 rdf:rest N124e9de582914687ae4ffe1fbbabd3e7
    103 anzsrc-for:08 schema:inDefinedTermSet anzsrc-for:
    104 schema:name Information and Computing Sciences
    105 rdf:type schema:DefinedTerm
    106 anzsrc-for:0801 schema:inDefinedTermSet anzsrc-for:
    107 schema:name Artificial Intelligence and Image Processing
    108 rdf:type schema:DefinedTerm
    109 sg:journal.1032807 schema:issn 0920-5691
    110 1573-1405
    111 schema:name International Journal of Computer Vision
    112 schema:publisher Springer Nature
    113 rdf:type schema:Periodical
    114 sg:person.010767212714.33 schema:affiliation grid-institutes:grid.1013.3
    115 schema:familyName Zhang
    116 schema:givenName Jing
    117 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.010767212714.33
    118 rdf:type schema:Person
    119 sg:person.014105751612.08 schema:affiliation grid-institutes:grid.1013.3
    120 schema:familyName Zhang
    121 schema:givenName Sen
    122 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.014105751612.08
    123 rdf:type schema:Person
    124 sg:person.0677067025.10 schema:affiliation grid-institutes:grid.1013.3
    125 schema:familyName Tao
    126 schema:givenName Dacheng
    127 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.0677067025.10
    128 rdf:type schema:Person
    129 sg:pub.10.1007/978-3-030-58536-5_24 schema:sameAs https://app.dimensions.ai/details/publication/pub.1132270164
    130 https://doi.org/10.1007/978-3-030-58536-5_24
    131 rdf:type schema:CreativeWork
    132 sg:pub.10.1007/s10462-012-9365-8 schema:sameAs https://app.dimensions.ai/details/publication/pub.1050263800
    133 https://doi.org/10.1007/s10462-012-9365-8
    134 rdf:type schema:CreativeWork
    135 sg:pub.10.1186/s41074-017-0027-2 schema:sameAs https://app.dimensions.ai/details/publication/pub.1085758120
    136 https://doi.org/10.1186/s41074-017-0027-2
    137 rdf:type schema:CreativeWork
    138 grid-institutes:grid.1013.3 schema:alternateName The University of Sydney, Sydney, NSW, Australia
    139 schema:name The University of Sydney, Sydney, NSW, Australia
    140 rdf:type schema:Organization
     




    Preview window. Press ESC to close (or click here)


    ...