Pranilllllll/geonusaf-segNext-block-fold0

GeoNUSAF - SegNeXt-T - block split, fold 0

Kathmandu Valley land-use segmentation, 6 classes, ignore_index=255.

field	value
architecture	MSCAN-T encoder + LightHamHead decoder (SegNeXt, NeurIPS 2022)
encoder init	ImageNet-1K, 100.0% of tensors loaded
params	4.23 M
head fuse stride	8 (stages [1, 2, 3])
NMF rank / steps	R=16, 6 train / 7 eval
split mode	block
fold	0 of 3
seed	42
input	512x512, ImageNet norm, effective GSD 0.586 m/px
classes	Residential, Road, River, Forest, UnusedLand, Agricultural
lr (head/encoder)	0.0006 / 6e-05
regularization	wd 0.01, drop_path 0.1, smooth 0.05, EMA True
best epoch	105
val mIoU	0.3345
val mF1	0.4615
val OA	0.6138
val kappa	0.4887

Per-class (validation)

class	IoU	F1
Residential	0.6138	0.7607
Road	0.2460	0.3949
River	0.0635	0.1195
Forest	0.5558	0.7144
UnusedLand	0.0980	0.1786
Agricultural	0.4297	0.6011

best.pt holds model_state (EMA weights when EMA is on), arch (the dict needed to rebuild the network), the run cfg, and metrics. Rebuild with segnext_model.py from this same repo. Model code derives from Visual-Attention-Network/SegNeXt (Apache-2.0).