RDAH-Net: Monocular Height Estimation Checkpoints for Remote Sensing

187.4 MB of model checkpoints and associated datasets for RDAH-Net, a method for generating normalized Digital Surface Models (nDSM) from single satellite images. The work, authored by Liting Jiang and published in 2026, uses a cross-modal fusion of depth priors and orthophotos to estimate absolute height. It is designed for large-scale mapping where stereo imagery is unavailable.

Use Cases

• Training monocular height estimation models based on the described RDAH-Net architecture.
• Generating normalized Digital Surface Models (nDSM) for large-scale mapping based on single satellite images.
• Evaluating cross-domain generalization performance for height recovery in remote sensing.
• Fusing relative depth priors with orthophoto data for robust absolute height estimation.

Strengths

• Model checkpoints are provided in a 187.4 MB package, enabling immediate use.
• The method is described as achieving strong cross-domain generalization for height recovery.
• The associated research is published in a peer-reviewed journal (MDPI Remote Sensing).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• The specific datasets included are not described in detail; their scale and content are unknown.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Liting Jiang via figshare, associated with research published in MDPI Remote Sensing.

Collection Method

Likely contains model weights (checkpoints) and datasets generated or used for the RDAH-Net research.

Time Range

null

Freshness

Last updated 2026-04-11 09:56:00; freshness should be verified.

Geography

null