GeoNUSAF TC-SegFormer (no_sampler) -- block split, fold 1

Kathmandu Valley land-use segmentation, 6 classes, ignore_index=255, 512x512 @ 0.586 m/px.

backbone nvidia/segformer-b0-finetuned-ade-512-512
split mode block (sequence-block CV, export-order proxy)
fold 1 of 3
ablation no_sampler
seed 42
best epoch 79
detail path True
CSA reweighting True (tau=[0.6, 0.35, 0.35, 0.6, 0.6, 0.6], w_min=0.25)
soft-clDice True (mu=0.3)
balanced sampler False

Validation

mIoU 0.5509 | mF1 0.6919 | OA 0.8255 | kappa 0.6846

class IoU UA (prec) PA (rec)
Residential 0.8471 0.9369 0.8983
Road 0.4307 0.5087 0.7376
River 0.4776 0.6874 0.6102
Forest 0.7031 0.9276 0.7440
UnusedLand 0.2769 0.5528 0.3568
Agricultural 0.5703 0.6210 0.8749
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support