Downloads · 30 days
0
Pranilllllll/geonusaf-s4-unet-r18-R1-block-fold1
geonusaf-s4-unet-r18-R1-block-fold1 is a machine learning model from Pranilllllll. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for segmentation-models-pytorch.
Kathmandu Valley land-use segmentation, 6 classes, ignoreindex=255. Weights are the EMA weights (decay 0.999), not the raw final weights.
Downloads · 30 days
0
Access
Public
Updated Sep 3, 2026
Repo size
3 GB
Likes
0
Public
Click a slice to open those files.
.pt250 MB · 98%
From the Hugging Face model README
Kathmandu Valley land-use segmentation, 6 classes, ignore_index=255. Weights are the EMA weights (decay 0.999), not the raw final weights.
| field | value |
|---|---|
| architecture | smp.Unet, encoder resnet18 (ImageNet), decoder [128, 64, 32, 16, 8] |
| params | 12.46 M |
| arch version | unet-r18-v1 |
| arm | R1 (real + 804 synthetic) |
| train pairs | 1608 (804 real + 804 synthetic) |
| synthetic source | sugam24/geonusaf-stage3-fakepairs-block-fold1, fake_pairs_R1 |
| split | block fold 1, train_sha1 eb0aebdd9919 |
| schedule | 6000 steps, warmup 500, cosine 6000 |
| class weights | from real |
| seed | 42 |
| best step | 3200 |
| val mIoU | 0.4515 |
| val mF1 | 0.5835 |
| val OA | 0.7738 |
| val kappa | 0.6217 |
| class | IoU | F1 |
|---|---|---|
| Residential | 0.8092 | 0.8945 |
| Road | 0.3142 | 0.4781 |
| River | 0.1181 | 0.2113 |
| Forest | 0.6622 | 0.7968 |
| UnusedLand | 0.2636 | 0.4173 |
| Agricultural | 0.5417 | 0.7027 |
Validation contains no synthetic pixels in either arm.
Not comparable to the part-1 fold-1 U-Net run: that one had frozen augmentation (the persistent_workers bug) and an epoch-based schedule.