Datasets:

anonymoussubmission2
/

anonymous-submission-neurips26-2831

layout_id stringdate 0004-01-01 00:00:00 0324-01-01 00:00:00	scenario_id stringlengths 5 5	season_number float32 2 4 ⌀	seasonal_match bool 1 class	historical_match bool 2 classes	big_fire bool 2 classes	small_fire bool 2 classes	fast_fire bool 2 classes	slow_fire bool 2 classes
0004	00225	2	false	false	true	false	false	true
0004	00640	2	false	false	true	false	false	true
0004	00591	2	false	false	true	false	false	true
0004	00993	2	false	false	true	false	false	true
0004	00732	2	false	false	true	false	false	true
0004	00304	2	false	false	true	false	false	true
0004	00590	2	false	false	true	false	false	true
0004	00584	2	false	false	true	false	false	true
0004	00583	2	false	false	true	false	false	true
0004	00945	2	false	false	true	false	true	false
0004	00582	2	false	false	true	false	false	true
0004	00296	2	false	false	true	false	false	true
0004	00652	2	false	false	true	false	false	true
0004	00644	2	false	false	true	false	false	true
0004	00592	2	false	false	true	false	false	true
0004	00205	2	false	false	true	false	false	true
0004	01174	2	false	false	true	false	false	true
0058	01359	null	false	false	true	false	false	true
0058	00088	null	false	false	true	false	false	true
0058	01961	null	false	false	true	false	false	true
0058	00017	null	false	false	true	false	false	true
0058	00213	null	false	false	true	false	false	true
0058	02312	null	false	false	true	false	false	true
0058	01960	null	false	false	true	false	false	true
0058	03062	null	false	false	true	false	false	true
0058	03091	null	false	false	true	false	false	true
0058	00018	null	false	false	true	false	false	true
0058	02313	null	false	false	true	false	false	true
0058	02914	null	false	false	true	false	false	true
0058	02913	null	false	false	true	false	false	true
0058	03712	null	false	false	true	false	false	true
0058	01841	null	false	false	true	false	false	true
0058	01246	null	false	false	true	false	false	true
0058	03542	null	false	false	true	false	false	true
0058	01020	null	false	false	true	false	false	true
0058	02007	null	false	false	true	false	false	true
0058	00554	null	false	false	true	false	false	true
0058	02832	null	false	false	true	false	false	true
0058	00553	null	false	false	true	false	false	true
0058	02452	null	false	false	true	false	false	true
0058	02009	null	false	false	true	false	false	true
0058	01840	null	false	false	true	false	false	true
0058	03714	null	false	false	true	false	false	true
0058	03713	null	false	false	true	false	false	true
0058	01222	null	false	false	true	false	false	true
0058	03544	null	false	false	true	false	false	true
0058	01019	null	false	false	true	false	false	true
0058	03308	null	false	false	true	false	false	true
0058	02681	null	false	false	true	false	false	true
0058	01838	null	false	false	true	false	false	true
0058	00122	null	false	false	true	false	false	true
0058	01467	null	false	false	true	false	false	true
0058	01469	null	false	false	true	false	false	true
0058	02471	null	false	false	true	false	false	true
0058	02015	null	false	false	true	false	false	true
0058	02046	null	false	false	true	false	false	true
0058	02070	null	false	false	true	false	false	true
0058	01601	null	false	false	true	false	false	true
0058	02470	null	false	false	true	false	false	true
0058	02014	null	false	false	true	false	false	true
0058	03132	null	false	false	true	false	false	true
0058	01837	null	false	false	true	false	false	true
0058	03307	null	false	false	true	false	false	true
0058	00576	null	false	false	true	false	false	true
0058	00578	null	false	false	true	false	false	true
0058	03309	null	false	false	true	false	false	true
0058	02680	null	false	false	true	false	false	true
0058	01839	null	false	false	true	false	false	true
0058	00345	null	false	false	true	false	false	true
0058	02071	null	false	false	true	false	false	true
0058	03226	null	false	false	true	false	false	true
0058	03448	null	false	false	true	false	false	true
0058	03228	null	false	false	true	false	false	true
0058	03446	null	false	false	true	false	false	true
0058	03076	null	false	false	true	false	false	true
0058	01973	null	false	false	true	false	false	true
0058	02962	null	false	false	true	false	false	true
0058	00652	null	false	false	true	false	false	true
0058	02595	null	false	false	true	false	false	true
0058	02963	null	false	false	true	false	false	true
0058	01972	null	false	false	true	false	false	true
0058	03077	null	false	false	true	false	false	true
0058	02327	null	false	false	true	false	false	true
0058	00219	null	false	false	true	false	false	true
0058	02549	null	false	false	true	false	false	true
0058	00687	null	false	false	true	false	false	true
0058	02547	null	false	false	true	false	false	true
0058	02974	null	false	false	true	false	false	true
0058	02973	null	false	false	true	false	false	true
0058	00627	null	false	false	true	false	false	true
0058	00686	null	false	false	true	false	false	true
0058	00220	null	false	false	true	false	false	true
0058	00628	null	false	false	true	false	false	true
0058	02838	null	false	false	true	false	false	true
0058	02467	null	false	false	true	false	false	true
0058	01220	null	false	false	true	false	false	true
0058	02831	null	false	false	true	false	false	true
0058	02469	null	false	false	true	false	false	true
0058	01842	null	false	false	true	false	false	true
0058	01619	null	false	false	true	false	false	true

End of preview. Expand in Data Studio

DroneBench Dataset (clean release)

Wildfire monitoring layouts for benchmarking sensor placement and drone routing (NeurIPS 2026 Datasets & Benchmarks anonymous submission).

Summary

Attribute	Value
Layouts in data archive	56 (49 + 7 tables23 extras)
Tables 2/3 split	471 scenarios / 12 layouts
Fire-spread	JPG in `Satellite_Images_Mask/`
Risk maps	`burn_map.npy` (oracle), `static_risk_bp2024.npy` (BP)

Large archives

DroneBench_data.tar — layouts (JPG + burn/BP + ancillary); no scenarii/*.npy

Loading indices

from datasets import load_dataset
REPO = "anonymoussubmission2/anonymous-submission-neurips26-2831"
ds = load_dataset(REPO, "default")       # full index
ds_t23 = load_dataset(REPO, "tables23")  # Tables 2/3 split

Selection rule for the Tables 2/3 split: layouts with ≥80% historical ignition match (splits/SELECTION_RULE.md).

Risk-map encoding (important)

static_risk_bp2024.npy — USFS Burn Probability (BP), the independent risk map used in Table 2. Stored as int16 with raw grid values typically in ~11–385 (not unit probabilities). These are FSim annual burn probability × 10 000 (USFS BP scaling). Convert to ([0,1]) with:
```
import numpy as np
bp = np.load("static_risk_bp2024.npy").astype(np.float32) / 10000.0
```
burn_map.npy — oracle / ground-truth dynamic risk map used in Table 3. Stored as float32 in [0, 1], shape (T, H, W) (mean burn fraction over scenarios at each time).

Legacy static_risk.npy is not shipped (it was a duplicate of BP under another name).

Fire-spread JPG encoding / thresholding

Scenarios live in Satellite_Images_Mask/<layout>_<scenario>/ as grayscale JPG frames (lossy).

Official preprocessing (code/dataset.py → load_scenario_jpg, used by preprocess_sim2real_dataset / jpg_scenario_to_npy):

Load each frame as grayscale (L).
Normalize: pixel / 255.0 → float in ([0,1]).
Binarize with threshold 0.5: (arr >= 0.5).astype(float).
Stack frames → NPY array shape (T, H, W).

JPEG compression can introduce mild artifacts; regenerating scenarii/*.npy with the library before experiments is recommended. Weather and other modalities in the release are unused by the current baseline strategies.

Ignition time offsets (`config_s2r.json`)

config_s2r.json maps offset_<layout>_<scenario> → integer hours to prepend empty frames when building burn maps / aligning scenarios. Coverage: all 7 746 scenarios in the default index and all 56 layouts (plus 7 single-scenario extras → 7 753 offset keys).

Citation

@inproceedings{anonymous_neurips2026_dronebench,
  title     = {A Benchmark for Sensor Placement and Drone Routing for Wildfire Detection},
  author    = {Anonymous},
  booktitle = {NeurIPS 2026 Datasets and Benchmarks Track},
  year      = {2026},
  note      = {Anonymous submission}
}

Downloads last month: 98