Datasets:
Add per-split sample index (JSONL) and pin dataset viewer to it
Browse files- .gitattributes +1 -0
- README.md +11 -2
- next_day_v3/splits/test.jsonl +0 -0
- next_day_v3/splits/train.jsonl +3 -0
- next_day_v3/splits/val.jsonl +0 -0
.gitattributes
CHANGED
|
@@ -59,3 +59,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 59 |
*.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 60 |
*.webm filter=lfs diff=lfs merge=lfs -text
|
| 61 |
next_day_v3/samples.json filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 59 |
*.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 60 |
*.webm filter=lfs diff=lfs merge=lfs -text
|
| 61 |
next_day_v3/samples.json filter=lfs diff=lfs merge=lfs -text
|
| 62 |
+
next_day_v3/splits/train.jsonl filter=lfs diff=lfs merge=lfs -text
|
README.md
CHANGED
|
@@ -8,7 +8,16 @@ tags:
|
|
| 8 |
- viirs
|
| 9 |
- earth-observation
|
| 10 |
size_categories:
|
| 11 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
---
|
| 13 |
|
| 14 |
# FireComp: Next-Day Fire Spread
|
|
@@ -24,8 +33,8 @@ Code, documentation, loaders and paper: **https://github.com/justuskarlsson/Fire
|
|
| 24 |
| Path | Description |
|
| 25 |
|---|---|
|
| 26 |
| `next_day_v3/` | Main dataset: 8 HDF5 shards (`dataset_*.h5`, zstd) + per-shard metadata (`dataset_*.json`), `samples.json`, `stats.json` (normalization) |
|
|
|
|
| 27 |
| `next_day_v3_case_study/` | Small case-study subset used for paper figures |
|
| 28 |
-
| `next_day_v3_test_multifire/` | Multi-fire test subset |
|
| 29 |
| `regions/` | Region definitions (`wildfire_regions.json` / `.tif`) |
|
| 30 |
| `fire_areas.npz` | Per-fire spatial area lookup (fire-size stratification) |
|
| 31 |
|
|
|
|
| 8 |
- viirs
|
| 9 |
- earth-observation
|
| 10 |
size_categories:
|
| 11 |
+
- 100K<n<1M
|
| 12 |
+
configs:
|
| 13 |
+
- config_name: default
|
| 14 |
+
data_files:
|
| 15 |
+
- split: train
|
| 16 |
+
path: next_day_v3/splits/train.jsonl
|
| 17 |
+
- split: validation
|
| 18 |
+
path: next_day_v3/splits/val.jsonl
|
| 19 |
+
- split: test
|
| 20 |
+
path: next_day_v3/splits/test.jsonl
|
| 21 |
---
|
| 22 |
|
| 23 |
# FireComp: Next-Day Fire Spread
|
|
|
|
| 33 |
| Path | Description |
|
| 34 |
|---|---|
|
| 35 |
| `next_day_v3/` | Main dataset: 8 HDF5 shards (`dataset_*.h5`, zstd) + per-shard metadata (`dataset_*.json`), `samples.json`, `stats.json` (normalization) |
|
| 36 |
+
| `next_day_v3/splits/{train,val,test}.jsonl` | Flat sample index per split (stratified temporal 60/15/25 within each region) — what the dataset viewer shows |
|
| 37 |
| `next_day_v3_case_study/` | Small case-study subset used for paper figures |
|
|
|
|
| 38 |
| `regions/` | Region definitions (`wildfire_regions.json` / `.tif`) |
|
| 39 |
| `fire_areas.npz` | Per-fire spatial area lookup (fire-size stratification) |
|
| 40 |
|
next_day_v3/splits/test.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
next_day_v3/splits/train.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ade5d2312b342140ee7becfc80b185e54f14110578a5fc189f95e1b6b282b4f0
|
| 3 |
+
size 14973462
|
next_day_v3/splits/val.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|