sugam24/geonusaf-tcsegformer-random-fold2

GeoNUSAF TC-SegFormer -- random split, fold 2

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	random (sequence-block CV, export-order proxy)
fold	2 of 3
seed	42
best epoch	110
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 0.5717 | mF1 0.7138 | OA 0.8189 | kappa 0.6937

class	IoU	UA (prec)	PA (rec)
Residential	0.8531	0.9237	0.9178
Road	0.4161	0.5516	0.6287
River	0.5710	0.7013	0.7545
Forest	0.7110	0.7831	0.8855
UnusedLand	0.3776	0.5406	0.5559
Agricultural	0.5016	0.6962	0.6422