Assessing the Performance of Deep Learning Networks for Real-Time Deforestation Segmentation in the Amazon and Atlantic Biomes
Year: 2025
Type: conference-paper
Abstract: Deforestation poses significant ecological and societal challenges, while advances in satellite imagery and deep learning have enhanced monitoring precision and scalability. This study evaluates ten deep learning models for deforestation segmentation, including U-Net, ResNet, FCN, and YOLO variants, assessing accuracy and computational efficiency, focusing on the Amazon and Atlantic Forest biomes. The results highlight that U-Net and ResNet50 achieve the highest accuracy, while YOLOv8 and YOLOv11 offer an optimal balance between speed and performance. The findings contribute to the model selection for real-time deforestation detection, supporting conservation and environmental decision-making in underexplored areas like the Atlantic Forest.
Institution: Universidade Estadual Paulista (Unesp)
Language: English
Subfield: Media Technology
Field: Engineering
Domain: Physical Sciences
SDGs: Life on land, Climate action
Open Access status: gold
Awards: Data Science for Smart Industry (CDII), G29316538, G3115900892, G4158202762, G5208889705, Enhancing Efficiency in Sugarcane Industry Forecasting through Integrated Data Analysis and Artificial Intelligence Techniques, Deep neural network-based approaches for change detection via remote sensing image series
OpenAlex ID: W4414226997