A Light-weight Deep Learning Model for Remote Sensing Image Classification

Lam Pham; Cam Le; Dat Ngo; Anh Nguyen; Jasmin Lampert; Alexander Schindler; Ian  McLoughlin

doi:10.1109/ISPA58351.2023.10279679

A Light-weight Deep Learning Model for Remote Sensing Image Classification

Lam Pham (Speaker, Invited), Cam Le (Author, Invited), Dat Ngo (Author, Invited), Anh Nguyen (Author, Invited), Jasmin Lampert (Author, Invited), Alexander Schindler (Author, Invited), Ian McLoughlin (Author, Invited)

Data Science & Artifical Intelligence

Research output: Chapter in Book or Conference Proceedings › Conference Proceedings with Oral Presentation › peer-review

Abstract

In this paper, we present a high-performance and light-weight deep learning model for Remote Sensing Image Classification (RSIC), the task of identifying the aerial scene of a remote sensing image. To this end, we first evaluate various benchmark convolutional neural network (CNN) architectures: MobileNet V1/V2, ResNet 50/151V2, InceptionV3/InceptionResNetV2, EfficientNet B0/B7, DenseNet 121/201, ConNeXt Tiny/Large. Then, the best performing models are selected to train a compact model in a teacher-student arrangement. The knowledge distillation from the teacher aims to achieve high performance with significantly reduced complexity. By conducting extensive experiments on the NWPU-RESISC45 benchmark, our proposed teacher and student models outper-forms the state-of-the-art systems, and has potential to be applied on a wide range of edge devices.

Original language	English
Title of host publication	13th Int'l Symposium on Image and Signal Processing and Analysis (ISPA 2023)
Number of pages	5
ISBN (Electronic)	979-8-3503-1536-3
DOIs	https://doi.org/10.1109/ISPA58351.2023.10279679
Publication status	Published - Jul 2023
Event	2023 International Symposium on Image and Signal Processing and Analysis (ISPA) - Rome, Rome, Italy Duration: 18 Sept 2023 → 19 Sept 2023

Conference

Conference	2023 International Symposium on Image and Signal Processing and Analysis (ISPA)
Country/Territory	Italy
City	Rome
Period	18/09/23 → 19/09/23

Research Field

Former Research Field - Data Science

Keywords

Teacher-student model
convolutional neural network (CNN)
data augmentation
high-level features

Access to Document

10.1109/ISPA58351.2023.10279679

Cite this

@inproceedings{fd10bdd5ad774045adf2a1816a6e6133,

title = "A Light-weight Deep Learning Model for Remote Sensing Image Classification",

abstract = "In this paper, we present a high-performance and light-weight deep learning model for Remote Sensing Image Classification (RSIC), the task of identifying the aerial scene of a remote sensing image. To this end, we first evaluate various benchmark convolutional neural network (CNN) architectures: MobileNet V1/V2, ResNet 50/151V2, InceptionV3/InceptionResNetV2, EfficientNet B0/B7, DenseNet 121/201, ConNeXt Tiny/Large. Then, the best performing models are selected to train a compact model in a teacher-student arrangement. The knowledge distillation from the teacher aims to achieve high performance with significantly reduced complexity. By conducting extensive experiments on the NWPU-RESISC45 benchmark, our proposed teacher and student models outper-forms the state-of-the-art systems, and has potential to be applied on a wide range of edge devices.",

keywords = "Teacher-student model, convolutional neural network (CNN), data augmentation, high-level features",

author = "Lam Pham and Cam Le and Dat Ngo and Anh Nguyen and Jasmin Lampert and Alexander Schindler and Ian McLoughlin",

year = "2023",

month = jul,

doi = "10.1109/ISPA58351.2023.10279679",

language = "English",

booktitle = "13th Int'l Symposium on Image and Signal Processing and Analysis (ISPA 2023)",

note = "2023 International Symposium on Image and Signal Processing and Analysis (ISPA) ; Conference date: 18-09-2023 Through 19-09-2023",

}

Pham, L, Le, C, Ngo, D, Nguyen, A, Lampert, J , Schindler, A & McLoughlin, I 2023, A Light-weight Deep Learning Model for Remote Sensing Image Classification. in 13th Int'l Symposium on Image and Signal Processing and Analysis (ISPA 2023). 2023 International Symposium on Image and Signal Processing and Analysis (ISPA), Rome, Italy, 18/09/23. https://doi.org/10.1109/ISPA58351.2023.10279679

A Light-weight Deep Learning Model for Remote Sensing Image Classification. / Pham, Lam (Speaker, Invited); Le, Cam (Author, Invited); Ngo, Dat (Author, Invited) et al.
13th Int'l Symposium on Image and Signal Processing and Analysis (ISPA 2023). 2023.

Research output: Chapter in Book or Conference Proceedings › Conference Proceedings with Oral Presentation › peer-review

TY - GEN

T1 - A Light-weight Deep Learning Model for Remote Sensing Image Classification

A2 - Pham, Lam

A2 - Le, Cam

A2 - Ngo, Dat

A2 - Nguyen, Anh

A2 - Lampert, Jasmin

A2 - Schindler, Alexander

A2 - McLoughlin, Ian

PY - 2023/7

Y1 - 2023/7

N2 - In this paper, we present a high-performance and light-weight deep learning model for Remote Sensing Image Classification (RSIC), the task of identifying the aerial scene of a remote sensing image. To this end, we first evaluate various benchmark convolutional neural network (CNN) architectures: MobileNet V1/V2, ResNet 50/151V2, InceptionV3/InceptionResNetV2, EfficientNet B0/B7, DenseNet 121/201, ConNeXt Tiny/Large. Then, the best performing models are selected to train a compact model in a teacher-student arrangement. The knowledge distillation from the teacher aims to achieve high performance with significantly reduced complexity. By conducting extensive experiments on the NWPU-RESISC45 benchmark, our proposed teacher and student models outper-forms the state-of-the-art systems, and has potential to be applied on a wide range of edge devices.

AB - In this paper, we present a high-performance and light-weight deep learning model for Remote Sensing Image Classification (RSIC), the task of identifying the aerial scene of a remote sensing image. To this end, we first evaluate various benchmark convolutional neural network (CNN) architectures: MobileNet V1/V2, ResNet 50/151V2, InceptionV3/InceptionResNetV2, EfficientNet B0/B7, DenseNet 121/201, ConNeXt Tiny/Large. Then, the best performing models are selected to train a compact model in a teacher-student arrangement. The knowledge distillation from the teacher aims to achieve high performance with significantly reduced complexity. By conducting extensive experiments on the NWPU-RESISC45 benchmark, our proposed teacher and student models outper-forms the state-of-the-art systems, and has potential to be applied on a wide range of edge devices.

KW - Teacher-student model

KW - convolutional neural network (CNN)

KW - data augmentation

KW - high-level features

U2 - 10.1109/ISPA58351.2023.10279679

DO - 10.1109/ISPA58351.2023.10279679

M3 - Conference Proceedings with Oral Presentation

BT - 13th Int'l Symposium on Image and Signal Processing and Analysis (ISPA 2023)

T2 - 2023 International Symposium on Image and Signal Processing and Analysis (ISPA)

Y2 - 18 September 2023 through 19 September 2023

ER -

A Light-weight Deep Learning Model for Remote Sensing Image Classification

Abstract

Conference

Research Field

Keywords

Access to Document

Fingerprint

Cite this