Image Segmentation
fastai
PyTorch
geospatial
remote-sensing
semantic-segmentation
sentinel-1
alphaearth
flood-mapping
surface-water
Instructions to use rohitm9/surfaceWaterGlobal_flood with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- fastai
How to use rohitm9/surfaceWaterGlobal_flood with fastai:
from huggingface_hub import from_pretrained_fastai learn = from_pretrained_fastai("rohitm9/surfaceWaterGlobal_flood") - Notebooks
- Google Colab
- Kaggle
Download MODEL_INFO.json from rohitm9/surfaceWaterGlobal_flood: direct link, hf CLI and curl.
- Browser
- Download file 2.38 kB
-
https://huggingface.co/rohitm9/surfaceWaterGlobal_flood/resolve/main/MODEL_INFO.json
- Command line
-
hf download hf://rohitm9/surfaceWaterGlobal_flood/MODEL_INFO.json
-
curl -L -o MODEL_INFO.json https://huggingface.co/rohitm9/surfaceWaterGlobal_flood/resolve/main/MODEL_INFO.json
2.38 kB
| { | |
| "model_id": "rohitm9/surfaceWaterGlobal_flood", | |
| "version": "1.1.0", | |
| "created": "2026-11-05", | |
| "task": "binary-semantic-segmentation (water incl. flood water)", | |
| "framework": [ | |
| "pytorch", | |
| "fastai" | |
| ], | |
| "architecture": "resnet34-unet-with-aef-bottleneck", | |
| "inputs": { | |
| "sentinel_1": { | |
| "bands": 3, | |
| "order": [ | |
| "VV", | |
| "VH", | |
| "angle" | |
| ], | |
| "source": "Earth Engine COPERNICUS/S1_GRD, dB, 10 m" | |
| }, | |
| "alphaearth": { | |
| "bands": 64, | |
| "order": "source order", | |
| "year_rule": "max(S1 year - 1, 2017)" | |
| } | |
| }, | |
| "aef_projection": { | |
| "input_channels": 64, | |
| "output_channels": 16, | |
| "kernel_size": 1 | |
| }, | |
| "aef_dropout_p": 0.0, | |
| "training": { | |
| "sets": "W (Sen1Floods11 weak labels) + O (global post-flood chips, Dynamic World label class all water)", | |
| "train_chips": 6994, | |
| "valid_chips": 504, | |
| "loss": "ce_dice", | |
| "sampling": "every 512 px tile of every chip once per epoch, area weighted", | |
| "checkpoint_selection": "source_balanced_water_iou, best over all stages" | |
| }, | |
| "normalization": "band_stats.npz next to each checkpoint (67 bands S1+AEF, 3 bands S1-only)", | |
| "evaluation_files": [ | |
| "evaluation/summary_all_tests.csv", | |
| "evaluation/per_seed.csv", | |
| "evaluation/per_event.csv", | |
| "evaluation/contrasts_event_bootstrap.csv" | |
| ], | |
| "evaluation_note": "evaluation/*.csv report means over 3 training seeds (42, 43, 44); only seed 42 of each model is distributed. Seed 43/44 thresholds: S1+AEF 0.05/0.10, S1-only 0.05/0.04.", | |
| "models": { | |
| "s1_aef": { | |
| "model_kind": "s1aef", | |
| "run_dir": "seed42", | |
| "inputs": "S1 (3) + AEF (64)", | |
| "internal_run_name": "full_coverage_v1/FC_AEF_WO", | |
| "checkpoint": "seed42/models/s1aef_bottleneck_resnet34_best.pth", | |
| "checkpoint_sha256": "cddb479fe11b24ca5f70d460924e5f35b640cbc2f5894a5f931f78bd8dddebe4", | |
| "validation_threshold": 0.1, | |
| "validation_water_iou_tta": 0.7359 | |
| }, | |
| "s1_only": { | |
| "model_kind": "s1", | |
| "run_dir": "s1_only/seed42", | |
| "inputs": "S1 (3)", | |
| "internal_run_name": "full_coverage_v1/FC_S1_WO", | |
| "checkpoint": "s1_only/seed42/models/s1dw_resnet34_best.pth", | |
| "checkpoint_sha256": "f553e8a6477c9b468b5b2b37dc4b771ca812123c3d420a788983f516728b68bd", | |
| "validation_threshold": 0.04, | |
| "validation_water_iou_tta": 0.7065 | |
| } | |
| } | |
| } |