Fast implementation of real-time fruit detection in apple orchards using deep learning

Hanwen Kang, Chao Chen

Research output: Contribution to journalArticleResearchpeer-review

1 Citation (Scopus)

Abstract

To perform robust and efficient fruit detection in orchards is challenging since there are a number of variances in the working environments. Recently, deep-learning have shown a promising performance in many visual-guided agriculture applications. However, deep-learning based approaches requires labelling on training data, which is a labour-intensive and time-consuming task. In this study, a fast implementation framework of a deep-learning based fruit detector for apple harvesting is developed. The developed framework comprises an auto label generation module and a deep-learning-based fruit detector ‘LedNet’. The Label Generation algorithm utilises the multi-scale pyramid and clustering classifier to assist fast labelling of training data. LedNet adopts feature pyramid network and atrous spatial pyramid pooling to improve the detection performance of the model. A light-weight backbone is also developed and utilised to improve computational efficiency. From the experimental results, LedNet achieves 0.821 and 0.853 on recall and accuracy on apple detection in orchards, and its weights size and inference time are 7.4 M and 28 ms, respectively. The experimental results show that LedNet can perform real-time apple detection in orchards robustly and efficiently.

Original languageEnglish
Article number105108
Number of pages10
JournalComputers and Electronics in Agriculture
Volume168
DOIs
Publication statusPublished - Jan 2020

Keywords

  • Data labelling
  • Deep learning
  • Fruit detection
  • Real-time
  • Robotic harvesting

Cite this

@article{017be9c2c0d74b8e828e9161e4b0a93b,
title = "Fast implementation of real-time fruit detection in apple orchards using deep learning",
abstract = "To perform robust and efficient fruit detection in orchards is challenging since there are a number of variances in the working environments. Recently, deep-learning have shown a promising performance in many visual-guided agriculture applications. However, deep-learning based approaches requires labelling on training data, which is a labour-intensive and time-consuming task. In this study, a fast implementation framework of a deep-learning based fruit detector for apple harvesting is developed. The developed framework comprises an auto label generation module and a deep-learning-based fruit detector ‘LedNet’. The Label Generation algorithm utilises the multi-scale pyramid and clustering classifier to assist fast labelling of training data. LedNet adopts feature pyramid network and atrous spatial pyramid pooling to improve the detection performance of the model. A light-weight backbone is also developed and utilised to improve computational efficiency. From the experimental results, LedNet achieves 0.821 and 0.853 on recall and accuracy on apple detection in orchards, and its weights size and inference time are 7.4 M and 28 ms, respectively. The experimental results show that LedNet can perform real-time apple detection in orchards robustly and efficiently.",
keywords = "Data labelling, Deep learning, Fruit detection, Real-time, Robotic harvesting",
author = "Hanwen Kang and Chao Chen",
year = "2020",
month = "1",
doi = "10.1016/j.compag.2019.105108",
language = "English",
volume = "168",
journal = "Computers and Electronics in Agriculture",
issn = "0168-1699",
publisher = "Elsevier",

}

Fast implementation of real-time fruit detection in apple orchards using deep learning. / Kang, Hanwen; Chen, Chao.

In: Computers and Electronics in Agriculture, Vol. 168, 105108, 01.2020.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Fast implementation of real-time fruit detection in apple orchards using deep learning

AU - Kang, Hanwen

AU - Chen, Chao

PY - 2020/1

Y1 - 2020/1

N2 - To perform robust and efficient fruit detection in orchards is challenging since there are a number of variances in the working environments. Recently, deep-learning have shown a promising performance in many visual-guided agriculture applications. However, deep-learning based approaches requires labelling on training data, which is a labour-intensive and time-consuming task. In this study, a fast implementation framework of a deep-learning based fruit detector for apple harvesting is developed. The developed framework comprises an auto label generation module and a deep-learning-based fruit detector ‘LedNet’. The Label Generation algorithm utilises the multi-scale pyramid and clustering classifier to assist fast labelling of training data. LedNet adopts feature pyramid network and atrous spatial pyramid pooling to improve the detection performance of the model. A light-weight backbone is also developed and utilised to improve computational efficiency. From the experimental results, LedNet achieves 0.821 and 0.853 on recall and accuracy on apple detection in orchards, and its weights size and inference time are 7.4 M and 28 ms, respectively. The experimental results show that LedNet can perform real-time apple detection in orchards robustly and efficiently.

AB - To perform robust and efficient fruit detection in orchards is challenging since there are a number of variances in the working environments. Recently, deep-learning have shown a promising performance in many visual-guided agriculture applications. However, deep-learning based approaches requires labelling on training data, which is a labour-intensive and time-consuming task. In this study, a fast implementation framework of a deep-learning based fruit detector for apple harvesting is developed. The developed framework comprises an auto label generation module and a deep-learning-based fruit detector ‘LedNet’. The Label Generation algorithm utilises the multi-scale pyramid and clustering classifier to assist fast labelling of training data. LedNet adopts feature pyramid network and atrous spatial pyramid pooling to improve the detection performance of the model. A light-weight backbone is also developed and utilised to improve computational efficiency. From the experimental results, LedNet achieves 0.821 and 0.853 on recall and accuracy on apple detection in orchards, and its weights size and inference time are 7.4 M and 28 ms, respectively. The experimental results show that LedNet can perform real-time apple detection in orchards robustly and efficiently.

KW - Data labelling

KW - Deep learning

KW - Fruit detection

KW - Real-time

KW - Robotic harvesting

UR - http://www.scopus.com/inward/record.url?scp=85076234578&partnerID=8YFLogxK

U2 - 10.1016/j.compag.2019.105108

DO - 10.1016/j.compag.2019.105108

M3 - Article

AN - SCOPUS:85076234578

VL - 168

JO - Computers and Electronics in Agriculture

JF - Computers and Electronics in Agriculture

SN - 0168-1699

M1 - 105108

ER -