Paper Accepted to SIBGRAPI 2022!
Our paper entitled “A High-Spatial Resolution Dataset and Few-shot Deep Learning Benchmark for Image Classification” was just accepted for publication in the SIBGRAPI - Conference on Graphics, Patterns and Images.
The Cerrado, the second largest Brazilian biome, covers 23% of the national territory, extending over two million square kilometers. It is characterized by typical phytophysiognomies in which woody plants have thick stems, a dark tone, and are twisted but, in other cases, the branches can be angled close to the ground and the tip facing upwards (Eiten, 1990). The term “Cerrado” has been used to refer to the biome, a set of vegetation physiognomies, as well as to a specific type of floristic composition that occurs in the formation of savannas (Ribeiro and Walter, 2008).
This biome is located in the tropical zone where climate indirectly affects the characteristics and development of vegetation through the soil (Eiten, 1990). For almost all the Cerrado, the climate is defined in two seasons: wet, occurring more frequently between September and April, and dry, occurring mainly between April and September. It is important to add that at least 40% of the entire area has been converted into pastures and extensive agricultural fields, specifically annual crops such as soybeans and corn (Reatto et al., 2008).
Hence, it is important to analyze the dynamic of land use and land cover (LULC) using methods, techniques, and also based on a robust dataset that offers diversity of data for each class. Several previous studies do not use high-spatial-resolution images set for classification using deep learning (DL) and remote sensing techniques for some Earth observation applications. Moreover, it is important that a significant number of images (tiles) is used in order to have a representative sample of a large biome like Cerrado.
In order to fulfill these gaps, we introduce a novel high-spatial-resolution dataset with optical remote sensing images of the Cerrado for LULC classification, aiming to facilitate access to data ready to support ML and DL models for classification and segmentation. The Biome Cerrado Dataset (CerraData) is a large database, a total of 2.5 million tiles of 256x256 pixels, obtained from 150 scenes made by the Wide Panchromatic and Multispectral Camera (WPM) of the China-Brazil Earth Resources-4A (CBERS-4A) satellite.
Are you looking for a robust dataset for studying the Brazilian Cerrado vegetation? Check our paper!
Jurandy Almeida
Professor of Computer Science
My research interests are mainly in the areas of computer vision, deep learning, image processing, information retrieval, machine learning, and pattern recognition.