Skip to content
← Directory

dataset · Global

snd-remask-dgle-remote-sensing-hpc

Hugging Face dataset kuan2/snd-remask-dgle-remote-sensing-hpc

Claim this listing

Represent snd-remask-dgle-remote-sensing-hpc? Send a claim for review so we can discuss listing updates. Claims do not automatically grant ownership or verification.

Explore paid sponsorship →

What's in it

Data offered docs-verified
Vaihingen IRRG labelled source, 256 patches (vh2pd) · Potsdam RGB unlabelled target train, 2,592 patches (vh2pd) · Potsdam RGB labelled target evaluation, 864 patches (vh2pd) · LoveDA Rural labelled source, 1,366 images (r2u) · LoveDA Urban unlabelled target train, 1,156 images (r2u) · LoveDA Urban Val labelled target evaluation, 677 images (r2u) · ImageNet ResNet101 initialization weights
Coverage docs-verified
Vaihingen and Potsdam (ISPRS benchmark imagery); LoveDA Rural and Urban scenes
Formats docs-verified
PNG · ZIP64 archive · CSV · PyTorch .pth (ResNet101 initialization)
Access docs-verified
Public Hugging Face dataset, no authentication required. Licence: LoveDA portion is CC BY-NC-SA 4.0 (academic, noncommercial use only); ISPRS imagery is covered by owner's permission to distribute, with no onward redistribution rights granted.

Useful for

Use cases llm-inferred
Benchmarking source-only and domain-adapted aerial segmentation models across Vaihingen/Potsdam and LoveDA domain shifts · Reproducing published source-only baseline IoU scores before running an adaptation experiment · Prototyping land-cover segmentation on aerial imagery where no local labelled data exists
Audience llm-inferred
remote sensing researchers · computer vision and ML engineers · GIS analysts working on research segmentation tasks

Limitations

Constraints llm-inferred
Research replication bundle for the DGLE protocol, not an operational or consultancy data feed · Replication is a documented attempt, not the authors' official code; some source-training details are unspecified · Target training split has no labels: no label paths or masks are shipped · Hugging Face dataset viewer is unavailable for the train split (FeaturesError) · No onward redistribution rights are granted for the bundle · Named imagery sites (Vaihingen, Potsdam, LoveDA) show no UK coverage; registry region 'Global' is broader than the docs support · Restore and experiments need CUDA-enabled PyTorch and a GPU allocation; total size 6.8 GB · Raw ISPRS/Zenodo archives are excluded; rebuilding from source needs the ISPRS share password
Licence
Unknown — check provider terms
Authentication
Unknown — review source docs
GitHub stars
Unknown
Last commit
Unknown
Last successful page check
14/10/2026
Page health
healthy
Package
Not recorded
Verification
Not independently verified

Provenance & evidence

Verified 2026-10-15 via HF API: exists; 322 downloads/30d, 0 likes.

Metadata observed 14/10/2026 · research sweep: manual-2026-10-15 (Hugging Face datasets)