Noppitak, S., Okafor, E., & Surinta, O. (2024). EcoCropsAID: Economic Crops Aerial Image Dataset for Land Use Classification. Data in Brief, 49, 109541. https://doi.org/10.1016/j.dib.2024.109541
This data descriptor introduces the EcoCropsAID dataset, aiming to provide a valuable resource for researchers developing and evaluating land use classification algorithms, particularly for economic crops in Thailand. The authors highlight the challenges posed by the dataset, encouraging the exploration of novel approaches in computer vision and deep learning.
The EcoCropsAID dataset was created by collecting 5,400 aerial images of five economic crops in Thailand (rice, sugarcane, cassava, rubber, and longan) using the Google Earth application between 2014 and 2018. The images were selected from various locations in northeastern Thailand, representing different crop growth stages and exhibiting variations in resolution, color, and contrast due to the use of multiple remote imaging sensors.
The EcoCropsAID dataset is characterized by significant intra-class variability and inter-class similarity, making accurate land use classification challenging. The authors emphasize the need for advanced algorithms capable of handling these complexities and suggest exploring spatial-temporal feature extraction, deep learning architectures, and transformer-based models.
The EcoCropsAID dataset offers a valuable platform for advancing research in land use classification, particularly for economic crops. The dataset's challenges encourage the development of novel algorithms that can effectively address issues related to image variability and inter-class similarity, potentially leading to improved accuracy and efficiency in land use analysis.
This research contributes to the field of computer vision and remote sensing by providing a publicly available dataset specifically designed to address the challenges of economic crop classification from aerial imagery. The findings highlight the need for sophisticated algorithms and encourage further research in this area.
The study is limited to five economic crops in northeastern Thailand. Future research could expand the dataset to include a wider variety of crops and geographical locations. Additionally, exploring the integration of spatial-temporal data and the development of hybrid deep learning models could further enhance classification accuracy.
To Another Language
from source content
arxiv.org
Key Insights Distilled From
by Sangdaow Nop... at arxiv.org 11-06-2024
https://arxiv.org/pdf/2411.02762.pdfDeeper Inquiries