IDENTIFICATION OF NATURAL OBJECTS USING DEEP LEARNING AND ADDITIONAL DATA PREPROCESSING
The classification of natural objects in the wild is a popular task in the field of tourism and remote sensing. The key problem is the requirement for system performance in the absence of Internet access and a small amount of available resources, such as a mobile phone. In this regard, to solve the...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Copernicus Publications
2023-05-01
|
| Series: | The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences |
| Online Access: | https://isprs-archives.copernicus.org/articles/XLVIII-2-W3-2023/95/2023/isprs-archives-XLVIII-2-W3-2023-95-2023.pdf |
| _version_ | 1797828392004878336 |
|---|---|
| author | D. A. Kalashnikova V. V. Buryachenko |
| author_facet | D. A. Kalashnikova V. V. Buryachenko |
| author_sort | D. A. Kalashnikova |
| collection | DOAJ |
| description | The classification of natural objects in the wild is a popular task in the field of tourism and remote sensing. The key problem is the requirement for system performance in the absence of Internet access and a small amount of available resources, such as a mobile phone. In this regard, to solve the classification problem, it is required to use fairly simple neural networks and rely on a small amount of training data. The paper presents an image preprocessing method for object recognition in the the “Stolby National Park” in Krasnoyarsk city using a neural network. The approach involves applying a set of methods to expand the original training set. To analyze the effectiveness, several different neural networks based on MobileNET V2 are used, which makes it possible to compare test results on the original and extended data sets. We also evaluate the quality of objects identification on open datasets, such as Animals-10 and Landscape Pictures. The results of the experiments show the efficiency of data preprocessing, as well as the high performance of the modified neural network structure for the task of classifying natural objects in the environment. |
| first_indexed | 2024-04-09T13:03:33Z |
| format | Article |
| id | doaj.art-5acd95ce5e404d20a08aff37cbf15e20 |
| institution | Directory Open Access Journal |
| issn | 1682-1750 2194-9034 |
| language | English |
| last_indexed | 2024-04-09T13:03:33Z |
| publishDate | 2023-05-01 |
| publisher | Copernicus Publications |
| record_format | Article |
| series | The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences |
| spelling | doaj.art-5acd95ce5e404d20a08aff37cbf15e202023-05-12T17:24:13ZengCopernicus PublicationsThe International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences1682-17502194-90342023-05-01XLVIII-2-W3-20239510110.5194/isprs-archives-XLVIII-2-W3-2023-95-2023IDENTIFICATION OF NATURAL OBJECTS USING DEEP LEARNING AND ADDITIONAL DATA PREPROCESSINGD. A. Kalashnikova0V. V. Buryachenko1Reshetnev Siberian State University of Science and Technology, Institute of Informatics and Telecommunications, 31, Krasnoyarsky Rabochy ave., Krasnoyarsk, 660037, Russian FederationReshetnev Siberian State University of Science and Technology, Institute of Informatics and Telecommunications, 31, Krasnoyarsky Rabochy ave., Krasnoyarsk, 660037, Russian FederationThe classification of natural objects in the wild is a popular task in the field of tourism and remote sensing. The key problem is the requirement for system performance in the absence of Internet access and a small amount of available resources, such as a mobile phone. In this regard, to solve the classification problem, it is required to use fairly simple neural networks and rely on a small amount of training data. The paper presents an image preprocessing method for object recognition in the the “Stolby National Park” in Krasnoyarsk city using a neural network. The approach involves applying a set of methods to expand the original training set. To analyze the effectiveness, several different neural networks based on MobileNET V2 are used, which makes it possible to compare test results on the original and extended data sets. We also evaluate the quality of objects identification on open datasets, such as Animals-10 and Landscape Pictures. The results of the experiments show the efficiency of data preprocessing, as well as the high performance of the modified neural network structure for the task of classifying natural objects in the environment.https://isprs-archives.copernicus.org/articles/XLVIII-2-W3-2023/95/2023/isprs-archives-XLVIII-2-W3-2023-95-2023.pdf |
| spellingShingle | D. A. Kalashnikova V. V. Buryachenko IDENTIFICATION OF NATURAL OBJECTS USING DEEP LEARNING AND ADDITIONAL DATA PREPROCESSING The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences |
| title | IDENTIFICATION OF NATURAL OBJECTS USING DEEP LEARNING AND ADDITIONAL DATA PREPROCESSING |
| title_full | IDENTIFICATION OF NATURAL OBJECTS USING DEEP LEARNING AND ADDITIONAL DATA PREPROCESSING |
| title_fullStr | IDENTIFICATION OF NATURAL OBJECTS USING DEEP LEARNING AND ADDITIONAL DATA PREPROCESSING |
| title_full_unstemmed | IDENTIFICATION OF NATURAL OBJECTS USING DEEP LEARNING AND ADDITIONAL DATA PREPROCESSING |
| title_short | IDENTIFICATION OF NATURAL OBJECTS USING DEEP LEARNING AND ADDITIONAL DATA PREPROCESSING |
| title_sort | identification of natural objects using deep learning and additional data preprocessing |
| url | https://isprs-archives.copernicus.org/articles/XLVIII-2-W3-2023/95/2023/isprs-archives-XLVIII-2-W3-2023-95-2023.pdf |
| work_keys_str_mv | AT dakalashnikova identificationofnaturalobjectsusingdeeplearningandadditionaldatapreprocessing AT vvburyachenko identificationofnaturalobjectsusingdeeplearningandadditionaldatapreprocessing |