Table 6 Q2000 model validation compared with other deep learning studies of the Maya area
From: Evaluating Broadscale Deep Learning for Maya Settlement Detection in G-LiHT Lidar
Project | Study area (km2) | Training dataset | Precision | Accuracy | Recall | F-1 score | Visualization | Model type |
|---|---|---|---|---|---|---|---|---|
This paper | 35,584 | 2,000 | 0.91 | 0.82 | 0.88 | 0.89 | Custom | YOLOv3 |
Character 2024 | 615 | 15,765 | 0.98 | Not given | 0.67 | 0.80 | SLRM | YOLOv3 |
Richards-Rissetto 2021 | 26 | 410 | 0.85 | 0.85 | 0.89 | 0.87 | Hillshade | Inception-v3 CNN |
Richards-Rissetto 2021 | 26 | 142 | 0.91 | 0.95 | 0.94 | 0.92 | 3D Point Cloud | 3D PointConv |
Somrak 2020 | 130 | 8,687 | Not given | 0.99 | Not given | Not given | VAT | VGG-19 |
Bundzel 2020 | 160 | 670 | 0.72 | 0.99 | 0.6 | 0.62 | DEM | U-Net/Mask-RCNN |