Publication detail

RGB images-driven recognition of grapevine varieties using a densely connected convolutional network

ŠKRABÁNEK, P. DOLEŽEL, P. MATOUŠEK, R.

English title

RGB images-driven recognition of grapevine varieties using a densely connected convolutional network

Type

journal article in Web of Science

Language

en

Original abstract

We present a pocket-size densely connected convolutional network (DenseNet) directed to classification of size-normalized colour images according to varieties of grapes captured in those images. We compare the DenseNet with three established small-size networks in terms of performance, inference time, and model size. We propose a data augmentation that we use in training the networks. We train and evaluate the networks on in-field images. The trained networks distinguish between seven grapevine varieties and background, where four and three varieties, respectively are of red and green grapes. Compared to the established networks, the DenseNet is characterized by near state-of-the-art performance, short inference time, and minimal model size. All these aspects qualify the network for real-time, mobile, and edge computing applications. The DenseNet opens possibilities for constructing affordable selective harvesters in accordance with agriculture 4.0.

English abstract

We present a pocket-size densely connected convolutional network (DenseNet) directed to classification of size-normalized colour images according to varieties of grapes captured in those images. We compare the DenseNet with three established small-size networks in terms of performance, inference time, and model size. We propose a data augmentation that we use in training the networks. We train and evaluate the networks on in-field images. The trained networks distinguish between seven grapevine varieties and background, where four and three varieties, respectively are of red and green grapes. Compared to the established networks, the DenseNet is characterized by near state-of-the-art performance, short inference time, and minimal model size. All these aspects qualify the network for real-time, mobile, and edge computing applications. The DenseNet opens possibilities for constructing affordable selective harvesters in accordance with agriculture 4.0.

Keywords in English

Recognition of grapevine varieties; Densely connected convolutional network; Data augmentation; In-field images; Edge-computing; Agricultural mechanization

Released

16.02.2022

ISSN

1367-0751

Number

jzac029

Pages from–to

1–16

Pages count

16

BIBTEX


@article{BUT176693,
  author="Pavel {Škrabánek} and Petr {Doležel} and Radomil {Matoušek},
  title="RGB images-driven recognition of grapevine varieties using a densely connected convolutional network",
  year="2022",
  number="jzac029",
  month="February",
  pages="1--16",
  issn="1367-0751"
}