You are here: Home → Proceedings
Document Actions

Citation and metadata

Back to the article.

Recommended citation

Mayershofer C, Fischer A, Fottner J (2021). Improving Visual Object Detection Using Synthetic Self-Training. Logistics Journal : Proceedings, Vol. 2021. (urn:nbn:de:0009-14-54288)

Download Citation

Endnote

%0 Journal Article
%T Improving Visual Object Detection Using Synthetic Self-Training
%A Mayershofer, Christopher
%A Fischer, Adrian
%A Fottner, Johannes
%J Logistics Journal : Proceedings
%D 2021
%V 2021
%N 17
%@ 2192-9084
%F mayershofer2021
%X The current era of supervised learning requires a large corpus of application-specific training data with ground-truth annotations. The creation of large annotated datasets however is a costly endeavor. Moreover, the availability of a large annotated set of training data cannot be guaranteed in certain domains. Self-training attempts to overcome these problems by using a set of labeled data and a potentially infinite pool of unlabeled data to train a model in a semi-supervised manner. Self-training however only works if the annotated data is sufficient for training a strong teacher model, which depending on the application domain, is not always available. In this work, we propose and formulate a simple extension to the self-training paradigm and refer to it as Synthetic Self-Training (SST). SST is able to overcome the aforementioned problem by incorporating synthetically generated images into the training process, therefore improving model performance. Specifically, we address the problem of object detection in a logistics environment and are able to improve the state-of-the-art detection performance on the LOCO dataset by 12% mAP0.5.
%L 620
%K Logistics
%K LogistikSynthetic Self-Training
%K Object Detection
%K Objekterkennung
%K Selbst-Training
%K Self-Training
%K Synthetic Data
%K Synthetische Daten
%K Synthetisches Selbst-Training
%R 10.2195/lj_Proc_mayershofer_en_202112_01
%U http://nbn-resolving.de/urn:nbn:de:0009-14-54288
%U http://dx.doi.org/10.2195/lj_Proc_mayershofer_en_202112_01
Download

Bibtex

@Article{mayershofer2021,
  author = 	"Mayershofer, Christopher
		and Fischer, Adrian
		and Fottner, Johannes",
  title = 	"Improving Visual Object Detection Using Synthetic Self-Training",
  journal = 	"Logistics Journal : Proceedings",
  year = 	"2021",
  volume = 	"2021",
  number = 	"17",
  keywords = 	"Logistics; LogistikSynthetic Self-Training; Object Detection; Objekterkennung; Selbst-Training; Self-Training; Synthetic Data; Synthetische Daten; Synthetisches Selbst-Training",
  abstract = 	"The current era of supervised learning requires a large corpus of application-specific training data with ground-truth annotations. The creation of large annotated datasets however is a costly endeavor. Moreover, the availability of a large annotated set of training data cannot be guaranteed in certain domains. Self-training attempts to overcome these problems by using a set of labeled data and a potentially infinite pool of unlabeled data to train a model in a semi-supervised manner. Self-training however only works if the annotated data is sufficient for training a strong teacher model, which depending on the application domain, is not always available. In this work, we propose and formulate a simple extension to the self-training paradigm and refer to it as Synthetic Self-Training (SST). SST is able to overcome the aforementioned problem by incorporating synthetically generated images into the training process, therefore improving model performance. Specifically, we address the problem of object detection in a logistics environment and are able to improve the state-of-the-art detection performance on the LOCO dataset by 12{\%} mAP0.5.",
  issn = 	"2192-9084",
  doi = 	"10.2195/lj_Proc_mayershofer_en_202112_01",
  url = 	"http://nbn-resolving.de/urn:nbn:de:0009-14-54288"
}
Download

RIS

TY  - JOUR
AU  - Mayershofer, Christopher
AU  - Fischer, Adrian
AU  - Fottner, Johannes
PY  - 2021
DA  - 2021//
TI  - Improving Visual Object Detection Using Synthetic Self-Training
JO  - Logistics Journal : Proceedings
VL  - 2021
IS  - 17
KW  - Logistics
KW  - LogistikSynthetic Self-Training
KW  - Object Detection
KW  - Objekterkennung
KW  - Selbst-Training
KW  - Self-Training
KW  - Synthetic Data
KW  - Synthetische Daten
KW  - Synthetisches Selbst-Training
AB  - The current era of supervised learning requires a large corpus of application-specific training data with ground-truth annotations. The creation of large annotated datasets however is a costly endeavor. Moreover, the availability of a large annotated set of training data cannot be guaranteed in certain domains. Self-training attempts to overcome these problems by using a set of labeled data and a potentially infinite pool of unlabeled data to train a model in a semi-supervised manner. Self-training however only works if the annotated data is sufficient for training a strong teacher model, which depending on the application domain, is not always available. In this work, we propose and formulate a simple extension to the self-training paradigm and refer to it as Synthetic Self-Training (SST). SST is able to overcome the aforementioned problem by incorporating synthetically generated images into the training process, therefore improving model performance. Specifically, we address the problem of object detection in a logistics environment and are able to improve the state-of-the-art detection performance on the LOCO dataset by 12% mAP0.5.
SN  - 2192-9084
UR  - http://nbn-resolving.de/urn:nbn:de:0009-14-54288
DO  - 10.2195/lj_Proc_mayershofer_en_202112_01
ID  - mayershofer2021
ER  -
Download

Wordbib

<?xml version="1.0" encoding="UTF-8"?>
<b:Sources SelectedStyle="" xmlns:b="http://schemas.openxmlformats.org/officeDocument/2006/bibliography"  xmlns="http://schemas.openxmlformats.org/officeDocument/2006/bibliography" >
<b:Source>
<b:Tag>mayershofer2021</b:Tag>
<b:SourceType>ArticleInAPeriodical</b:SourceType>
<b:Year>2021</b:Year>
<b:PeriodicalTitle>Logistics Journal : Proceedings</b:PeriodicalTitle>
<b:Volume>2021</b:Volume>
<b:Issue>17</b:Issue>
<b:Url>http://nbn-resolving.de/urn:nbn:de:0009-14-54288</b:Url>
<b:Url>http://dx.doi.org/10.2195/lj_Proc_mayershofer_en_202112_01</b:Url>
<b:Author>
<b:Author><b:NameList>
<b:Person><b:Last>Mayershofer</b:Last><b:First>Christopher</b:First></b:Person>
<b:Person><b:Last>Fischer</b:Last><b:First>Adrian</b:First></b:Person>
<b:Person><b:Last>Fottner</b:Last><b:First>Johannes</b:First></b:Person>
</b:NameList></b:Author>
</b:Author>
<b:Title>Improving Visual Object Detection Using Synthetic Self-Training</b:Title>
<b:Comments>The current era of supervised learning requires a large corpus of application-specific training data with ground-truth annotations. The creation of large annotated datasets however is a costly endeavor. Moreover, the availability of a large annotated set of training data cannot be guaranteed in certain domains. Self-training attempts to overcome these problems by using a set of labeled data and a potentially infinite pool of unlabeled data to train a model in a semi-supervised manner. Self-training however only works if the annotated data is sufficient for training a strong teacher model, which depending on the application domain, is not always available. In this work, we propose and formulate a simple extension to the self-training paradigm and refer to it as Synthetic Self-Training (SST). SST is able to overcome the aforementioned problem by incorporating synthetically generated images into the training process, therefore improving model performance. Specifically, we address the problem of object detection in a logistics environment and are able to improve the state-of-the-art detection performance on the LOCO dataset by 12% mAP0.5.</b:Comments>
</b:Source>
</b:Sources>
Download

ISI

PT Journal
AU Mayershofer, C
   Fischer, A
   Fottner, J
TI Improving Visual Object Detection Using Synthetic Self-Training
SO Logistics Journal : Proceedings
PY 2021
VL 2021
IS 17
DI 10.2195/lj_Proc_mayershofer_en_202112_01
DE Logistics; LogistikSynthetic Self-Training; Object Detection; Objekterkennung; Selbst-Training; Self-Training; Synthetic Data; Synthetische Daten; Synthetisches Selbst-Training
AB The current era of supervised learning requires a large corpus of application-specific training data with ground-truth annotations. The creation of large annotated datasets however is a costly endeavor. Moreover, the availability of a large annotated set of training data cannot be guaranteed in certain domains. Self-training attempts to overcome these problems by using a set of labeled data and a potentially infinite pool of unlabeled data to train a model in a semi-supervised manner. Self-training however only works if the annotated data is sufficient for training a strong teacher model, which depending on the application domain, is not always available. In this work, we propose and formulate a simple extension to the self-training paradigm and refer to it as Synthetic Self-Training (SST). SST is able to overcome the aforementioned problem by incorporating synthetically generated images into the training process, therefore improving model performance. Specifically, we address the problem of object detection in a logistics environment and are able to improve the state-of-the-art detection performance on the LOCO dataset by 12% mAP0.5.
ER
Download

Mods

<mods>
  <titleInfo>
    <title>Improving Visual Object Detection Using Synthetic Self-Training</title>
  </titleInfo>
  <name type="personal">
    <namePart type="family">Mayershofer</namePart>
    <namePart type="given">Christopher</namePart>
  </name>
  <name type="personal">
    <namePart type="family">Fischer</namePart>
    <namePart type="given">Adrian</namePart>
  </name>
  <name type="personal">
    <namePart type="family">Fottner</namePart>
    <namePart type="given">Johannes</namePart>
  </name>
  <abstract>The current era of supervised learning requires a large corpus of application-specific training data with ground-truth annotations. The creation of large annotated datasets however is a costly endeavor. Moreover, the availability of a large annotated set of training data cannot be guaranteed in certain domains. Self-training attempts to overcome these problems by using a set of labeled data and a potentially infinite pool of unlabeled data to train a model in a semi-supervised manner. Self-training however only works if the annotated data is sufficient for training a strong teacher model, which depending on the application domain, is not always available. In this work, we propose and formulate a simple extension to the self-training paradigm and refer to it as Synthetic Self-Training (SST). SST is able to overcome the aforementioned problem by incorporating synthetically generated images into the training process, therefore improving model performance. Specifically, we address the problem of object detection in a logistics environment and are able to improve the state-of-the-art detection performance on the LOCO dataset by 12% mAP0.5.</abstract>
  <subject>
    <topic>Logistics</topic>
    <topic>LogistikSynthetic Self-Training</topic>
    <topic>Object Detection</topic>
    <topic>Objekterkennung</topic>
    <topic>Selbst-Training</topic>
    <topic>Self-Training</topic>
    <topic>Synthetic Data</topic>
    <topic>Synthetische Daten</topic>
    <topic>Synthetisches Selbst-Training</topic>
  </subject>
  <classification authority="ddc">620</classification>
  <relatedItem type="host">
    <genre authority="marcgt">periodical</genre>
    <genre>academic journal</genre>
    <titleInfo>
      <title>Logistics Journal : Proceedings</title>
    </titleInfo>
    <part>
      <detail type="volume">
        <number>2021</number>
      </detail>
      <detail type="issue">
        <number>17</number>
      </detail>
      <date>2021</date>
    </part>
  </relatedItem>
  <identifier type="issn">2192-9084</identifier>
  <identifier type="urn">urn:nbn:de:0009-14-54288</identifier>
  <identifier type="doi">10.2195/lj_Proc_mayershofer_en_202112_01</identifier>
  <identifier type="uri">http://nbn-resolving.de/urn:nbn:de:0009-14-54288</identifier>
  <identifier type="citekey">mayershofer2021</identifier>
</mods>
Download

Full Metadata

Bibliographic Citation Logistics Journal : referierte Veröffentlichungen, Vol. 2021, Iss. 17
Title

Improving Visual Object Detection Using Synthetic Self-Training (eng)
Author Christopher Mayershofer, Adrian Fischer, Johannes Fottner
Language eng
Abstract The current era of supervised learning requires a large corpus of application-specific training data with ground-truth annotations. The creation of large annotated datasets however is a costly endeavor. Moreover, the availability of a large annotated set of training data cannot be guaranteed in certain domains. Self-training attempts to overcome these problems by using a set of labeled data and a potentially infinite pool of unlabeled data to train a model in a semi-supervised manner. Self-training however only works if the annotated data is sufficient for training a strong teacher model, which depending on the application domain, is not always available. In this work, we propose and formulate a simple extension to the self-training paradigm and refer to it as Synthetic Self-Training (SST). SST is able to overcome the aforementioned problem by incorporating synthetically generated images into the training process, therefore improving model performance. Specifically, we address the problem of object detection in a logistics environment and are able to improve the state-of-the-art detection performance on the LOCO dataset by 12% mAP0.5. Die gegenwärtige Praxis des überwachten Lernens erfordert einen umfangreichen Korpus annotierter Trainingsdaten. Die Erstellung großer annotierter Datensätze ist jedoch ein kostspieliges Unterfangen. Darüber hinaus variiert die Verfügbarkeit eines großen annotierten Trainingsdatensatzes über unterschiedliche Anwendungsbereiche. Selbsttraining versucht, diese Probleme zu überwinden, indem eine Kombination aus annotierten Daten und nicht-annotierten Daten verwendet wird, um ein Modell zu trainieren. Selbsttraining bedarf jedoch einer ausreichenden Menge annotierter Trainingsdaten, um ein starkes Lehrermodell zu trainieren. Diese Arbeit stellt das Synthetische Selbst-Training (SST) vor, eine Erweiterung des konventionellen Selbst-Trainings. SST löst zuvor genannte Problem, durch Einbeziehung synthetisch erzeugte Daten in den Trainingsprozess. Diese Arbeit formuliert SST im Bereich der Visuellen Objekterkennung und zeigt empirische dessen Vorteile. Konkret ermöglicht es SST die Erkennungsgenauigkeit logistikspezifischer Objekte im LOCO Benchmark um 12% mAP0.5 zu verbessern.
Subject Logistics, LogistikSynthetic Self-Training, Object Detection, Objekterkennung, Selbst-Training, Self-Training, Synthetic Data, Synthetische Daten, Synthetisches Selbst-Training
DDC 620
Rights fDPPL
URN: urn:nbn:de:0009-14-54288
DOI https://doi.org/10.2195/lj_Proc_mayershofer_en_202112_01