🫧 Has more instances (>999) than what can be shown here:
- http://purl.org/aida/BiAU-Net%...fire%20burnt%20area%20mapping. BiAU-Net achieved improvements over the Fire_cci51 baseline of 11.56% in Overall Accuracy, 29.08% in Precision, 7.06% in Recall, 19.90% in F1-score, 15.44% in Balanced Accuracy, 29.90% in Kappa Coefficient, and 28.29% in Matthews Correlation Coefficient for wildfire burnt area mapping. ( )
- http://purl.org/aida/The%20BiA...ire%20event%20in%20California. The BiAU-Net model demonstrated good generalizability across five testing areas in different continents (United States, Spain, Australia, Indonesia, and Kenya) when trained on a single wildfire event in California. ( )
- http://purl.org/aida/Bi-tempor...single-input%20U-Net%20models. Bi-temporal input incorporating both pre-fire and post-fire Sentinel-2 imagery enhanced model performance across diverse environmental areas compared to traditional single-input U-Net models. ( )
- http://purl.org/aida/The%20com...the%20Chico%20wildfire%20case. The combination of Dice Loss and Focal Loss as the loss function significantly improved model performance in handling imbalanced datasets, with BiAU-Net achieving a Balanced Accuracy score of 0.965 compared to 0.930 for BiU-Net in the Chico wildfire case. ( )
- http://purl.org/aida/Attention...xels%20are%20mixed%20together. Attention mechanisms in the U-Net architecture enabled the model to focus on burnt areas and improve accuracy and efficiency, especially in detecting edges and small areas where burnt and non-burnt pixels are mixed together. ( )
- http://purl.org/aida/The%20BiA...lation%20Coefficient%20values. The BiAU-Net model achieved the highest overall performance compared to state-of-the-art wildfire burnt area detection models, as evidenced by the highest F1-score and Matthews Correlation Coefficient values. ( )
- http://purl.org/aida/BiAU-Net%...en%20overlook%20small%20fires. BiAU-Net using Sentinel-2 imagery at 20 m spatial resolution offers higher accuracy than global wildfire products Fire_cci51 and MODIS MCD64A1, which have limitations at 250 m and 500 m resolution and often overlook small fires. ( )
- http://purl.org/aida/The%20BiA...20%20m%20spatial%20resolution. The BiAU-Net model was trained using Sentinel-2 MSI Level 2A product with a false color band combination of Band12 (SWIR: 2114.9-2289.9 μm), Band11 (SWIR: 1568.2-1659.2 μm), and Band8A (Narrow NIR: 854.2-875.2 μm) at 20 m spatial resolution. ( )
- http://purl.org/aida/For%20the...on%20Coefficient%20of%200.929. For the Chico, California test area, BiAU-Net with DLoss+FLoss achieved an Overall Accuracy of 0.968, Precision of 0.974, Recall of 0.949, F1-score of 0.961, Balanced Accuracy of 0.965, Kappa Coefficient of 0.934, and Matthews Correlation Coefficient of 0.929. ( )
- http://purl.org/aida/The%20mod...radient%20Descent%20optimizer. The model training utilized a batch size of 6, conducted 50 epochs, initialized the learning rate at 0.1 with scheduled decrease by a factor of 0.5 between the 30th to 50th epochs, and used Stochastic Gradient Descent optimizer. ( )
📝 Related Templates
- Expressing a statement about research as an AIDA sentence attributed to Leiden University - by Tobias Kuhn, 2026-01-14
- Expressing a statement about research as an AIDA sentence - by Alessa An Gambardella, 2025-06-06
- Expressing a statement as an AIDA sentence - by Tobias Kuhn, 2024-10-22
(none)