TG-GAN: Multi-Rate DEM Terrain Generalization Model Outputs

TG-GAN is a deep learning model for multi-rate Digital Elevation Model generalization, constrained by terrain morphological factors. The model was developed by 岩晨 万 and last updated on 2026-05-27. Experiments were conducted in mountainous regions of Chongqing, China, Alaska, and Colorado with downscaling factors from 2× to 5×.

Use Cases

• Generate multi-scale DEMs for cartographic representation based on the described 'structure-first' principle.
• Enhance terrain generalization for geomorphological analysis based on the integration of local relief and gradient factors.
• Compare terrain simplification methods based on the reported performance metrics like MAE, RMSE, and SSIM.

Strengths

• Model outputs are validated across three distinct mountainous regions: Chongqing, Alaska, and Colorado.
• Performance is evaluated against traditional interpolation and CNN-based methods using MAE, RMSE, and SSIM metrics.
• The dataset is 151.2 MB in size and shared under a CC-BY-4.0 license.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The description metadata is detailed but specific file contents and formats within the ZIP are not enumerated.

Provenance

Source

figshare

Collection Method

Outputs from the TG-GAN deep learning model, as described.

Freshness

Last updated 2026-05-27 20:25:24; freshness should be verified.

Geography

Mountainous regions of Chongqing (China), Alaska (USA), and Colorado (USA).