GeoNUSAF TC-SegFormer -- 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
seed 42
best epoch 116
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 True

Validation

mIoU 0.5582 | mF1 0.6999 | OA 0.8281 | kappa 0.6913

class IoU UA (prec) PA (rec)
Residential 0.8450 0.9408 0.8925
Road 0.4382 0.5229 0.7302
River 0.4806 0.6788 0.6221
Forest 0.6996 0.9132 0.7495
UnusedLand 0.3031 0.5644 0.3957
Agricultural 0.5826 0.6288 0.8882
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