Research Article
True Orthoimage Generation Using Airborne LiDAR Data with Generative Adversarial Network-Based Deep Learning Model
Table 2
Evaluation of training performance for Vaihingen test datasets.
| Experiment | Training data type | Tile size (pixel) | Overlap (%) | FID | SSIM |
| 1 | Intensity | | 50 | 353.18 | 0.371 | 2 | DSM | | 50 | 231.74 | 0.401 | 3 | Color label | | 50 | 222.51 | 0.379 | 4 | B/W label | | 50 | 194.76 | 0.383 | 5 | Intensity+DSM | | 50 | 250.44 | 0.396 | 6 | Intensity+B/W label | | 50 | 218.01 | 0.421 | 7 | DSM+B/W label | | 50 | 214.77 | 0.431 | 8 | Intensity+DSM+B/W label | | 50 | 131.81 | 0.434 | 9 | DSM | | 60 | 197.63 | 0.427 | 10 | DSM | | 50 | 258.59 | 0.395 | 11 | DSM | | 70 | 174.79 | 0.406 | 12 | DSM | | 50 | 369.21 | 0.368 | 13 | DSM | | 80 | 148.41 | 0.377 | 14 | DSM+B/W label | | 80 | 132.35 | 0.436 | 15 | Intensity+DSM+B/W label | | 80 | 117.67 | 0.505 |
|
|