A method for pre-sorting mixed mail using convolutional neural networks and transfer learning

Michael Schadler; Dominik Stadlthanner; Bastian Mayer; Michael Schedler; Christian Landschützer

A method for pre-sorting mixed mail using convolutional neural networks and transfer learning

Michael Schadler, Dominik Stadlthanner, Bastian Mayer, Michael Schedler, Christian Landschützer

Institute of Logistics Engineering (3090)

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

Abstract

Sorting mixed mail has become increasingly important for postal service
providers. Low-value consignments, that have unfavorable material
properties, such as odd shape or flexibility, have increased in terms of
volume in recent years and cause substantial problems in sorting
processes. In order to automate the sorting of mixed-mail items the so-
called ‘Free-Fall-Sorter’ was developed at the Institute of Logistics
Engineering at Graz University of Technology. In this paper, an image
classification system is proposed to pre-sort mixed mail for this new
machine. The computer vision system used for the experiments applies
transfer learning to a publicly available convolutional neural network
called ‘MobileNet V2’. The classification system is designed to be
composed of low-cost hardware components, yet capable of providing
real-time inference. The results show a high accuracy of the re-trained
model.

Translated title of the contribution	Ein Verfahren zur Vorsortierung von Postsendungen mittels faltenden neuronalen Netzen und Transferlernen
Original language	English
Title of host publication	MHCL 2022
Subtitle of host publication	XXIV International Conference on "Material Handling, Constructions and Logistics" September 21st - 23rd, 2022, Belgrade
Place of Publication	Belgrade
Publisher	Faculty of Mechanical Engineering, Belgrade University
Pages	71-80
Number of pages	10
ISBN (Print)	978-86-6060-134-8
Publication status	Published - Sept 2022
Event	24th International Conference on Material Handling, Contructions and Logistics: MHCL 2022 - Faculty of mechanical Engineering - University Belgrade, Belgrad, Serbia Duration: 21 Sept 2022 → 23 Sept 2022 https://www.mhcl.info

Conference

Conference	24th International Conference on Material Handling, Contructions and Logistics
Country/Territory	Serbia
City	Belgrad
Period	21/09/22 → 23/09/22
Other	XXIV International Conference on Material Handling, Contructions and Logistics
Internet address	https://www.mhcl.info

Fields of Expertise

Mobility & Production

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

A method for pre-sorting mixed mail using convolutional neural networks and transfer learning
Michael Schadler (Speaker)
21 Sept 2022 → 23 Sept 2022
Activity: Talk or presentation › Talk at conference or symposium › Science to science

Cite this

Schadler, M., Stadlthanner, D., Mayer, B., Schedler, M., & Landschützer, C. (2022). A method for pre-sorting mixed mail using convolutional neural networks and transfer learning. In MHCL 2022: XXIV International Conference on "Material Handling, Constructions and Logistics" September 21st - 23rd, 2022, Belgrade (pp. 71-80). Faculty of Mechanical Engineering, Belgrade University.

A method for pre-sorting mixed mail using convolutional neural networks and transfer learning. / Schadler, Michael ; Stadlthanner, Dominik; Mayer, Bastian et al.
MHCL 2022: XXIV International Conference on "Material Handling, Constructions and Logistics" September 21st - 23rd, 2022, Belgrade. Belgrade: Faculty of Mechanical Engineering, Belgrade University, 2022. p. 71-80.

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

Schadler, M , Stadlthanner, D, Mayer, B, Schedler, M & Landschützer, C 2022, A method for pre-sorting mixed mail using convolutional neural networks and transfer learning. in MHCL 2022: XXIV International Conference on "Material Handling, Constructions and Logistics" September 21st - 23rd, 2022, Belgrade. Faculty of Mechanical Engineering, Belgrade University, Belgrade, pp. 71-80, 24th International Conference on Material Handling, Contructions and Logistics, Belgrad, Serbia, 21/09/22.

Schadler M , Stadlthanner D, Mayer B, Schedler M, Landschützer C. A method for pre-sorting mixed mail using convolutional neural networks and transfer learning. In MHCL 2022: XXIV International Conference on "Material Handling, Constructions and Logistics" September 21st - 23rd, 2022, Belgrade. Belgrade: Faculty of Mechanical Engineering, Belgrade University. 2022. p. 71-80

Schadler, Michael ; Stadlthanner, Dominik ; Mayer, Bastian et al. / A method for pre-sorting mixed mail using convolutional neural networks and transfer learning. MHCL 2022: XXIV International Conference on "Material Handling, Constructions and Logistics" September 21st - 23rd, 2022, Belgrade. Belgrade : Faculty of Mechanical Engineering, Belgrade University, 2022. pp. 71-80

@inproceedings{11fac3947aee45759b40882b4fea89f2,

title = "A method for pre-sorting mixed mail using convolutional neural networks and transfer learning",

abstract = "Sorting mixed mail has become increasingly important for postal serviceproviders. Low-value consignments, that have unfavorable materialproperties, such as odd shape or flexibility, have increased in terms ofvolume in recent years and cause substantial problems in sortingprocesses. In order to automate the sorting of mixed-mail items the so-called {\textquoteleft}Free-Fall-Sorter{\textquoteright} was developed at the Institute of LogisticsEngineering at Graz University of Technology. In this paper, an imageclassification system is proposed to pre-sort mixed mail for this newmachine. The computer vision system used for the experiments appliestransfer learning to a publicly available convolutional neural networkcalled {\textquoteleft}MobileNet V2{\textquoteright}. The classification system is designed to becomposed of low-cost hardware components, yet capable of providingreal-time inference. The results show a high accuracy of the re-trainedmodel.",

keywords = "polybag, Convolutional neural networks (CNNs), image classification, deep learning, low-value consignment",

author = "Michael Schadler and Dominik Stadlthanner and Bastian Mayer and Michael Schedler and Christian Landsch{\"u}tzer",

year = "2022",

month = sep,

language = "English",

isbn = "978-86-6060-134-8",

pages = "71--80",

booktitle = "MHCL 2022",

publisher = "Faculty of Mechanical Engineering, Belgrade University",

address = "Serbia",

note = "24th International Conference on Material Handling, Contructions and Logistics : MHCL 2022 ; Conference date: 21-09-2022 Through 23-09-2022",

url = "https://www.mhcl.info",

}

TY - GEN

T1 - A method for pre-sorting mixed mail using convolutional neural networks and transfer learning

AU - Schadler, Michael

AU - Stadlthanner, Dominik

AU - Mayer, Bastian

AU - Schedler, Michael

AU - Landschützer, Christian

PY - 2022/9

Y1 - 2022/9

N2 - Sorting mixed mail has become increasingly important for postal serviceproviders. Low-value consignments, that have unfavorable materialproperties, such as odd shape or flexibility, have increased in terms ofvolume in recent years and cause substantial problems in sortingprocesses. In order to automate the sorting of mixed-mail items the so-called ‘Free-Fall-Sorter’ was developed at the Institute of LogisticsEngineering at Graz University of Technology. In this paper, an imageclassification system is proposed to pre-sort mixed mail for this newmachine. The computer vision system used for the experiments appliestransfer learning to a publicly available convolutional neural networkcalled ‘MobileNet V2’. The classification system is designed to becomposed of low-cost hardware components, yet capable of providingreal-time inference. The results show a high accuracy of the re-trainedmodel.

AB - Sorting mixed mail has become increasingly important for postal serviceproviders. Low-value consignments, that have unfavorable materialproperties, such as odd shape or flexibility, have increased in terms ofvolume in recent years and cause substantial problems in sortingprocesses. In order to automate the sorting of mixed-mail items the so-called ‘Free-Fall-Sorter’ was developed at the Institute of LogisticsEngineering at Graz University of Technology. In this paper, an imageclassification system is proposed to pre-sort mixed mail for this newmachine. The computer vision system used for the experiments appliestransfer learning to a publicly available convolutional neural networkcalled ‘MobileNet V2’. The classification system is designed to becomposed of low-cost hardware components, yet capable of providingreal-time inference. The results show a high accuracy of the re-trainedmodel.

KW - polybag

KW - Convolutional neural networks (CNNs)

KW - image classification

KW - deep learning

KW - low-value consignment

M3 - Conference paper

SN - 978-86-6060-134-8

SP - 71

EP - 80

BT - MHCL 2022

PB - Faculty of Mechanical Engineering, Belgrade University

CY - Belgrade

T2 - 24th International Conference on Material Handling, Contructions and Logistics

Y2 - 21 September 2022 through 23 September 2022

ER -

A method for pre-sorting mixed mail using convolutional neural networks and transfer learning

Abstract

Conference

Fields of Expertise

UN SDGs

Activities

A method for pre-sorting mixed mail using convolutional neural networks and transfer learning

Cite this