File size: 2,161 Bytes
ed4b727 1b9bc05 9cbd7b2 1b9bc05 9cbd7b2 1b9bc05 9cbd7b2 ed4b727 1b9bc05 9cbd7b2 1b9bc05 9cbd7b2 1b9bc05 9cbd7b2 1b9bc05 9cbd7b2 1b9bc05 9cbd7b2 1b9bc05 9cbd7b2 1b9bc05 9cbd7b2 1b9bc05 9cbd7b2 1b9bc05 9cbd7b2 1b9bc05 9cbd7b2 1b9bc05 9cbd7b2 1b9bc05 9cbd7b2 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 |
---
datasets:
- ZoneTwelve/Thermal-Heatmap-Source-Localization
tags:
- benchmark
- heatmap
- physics
- source-localization
- synthetic
license: apache-2.0
pretty_name: Thermal Heatmap Source Localization (ThermBench)
---
# ThermBench 🔥 — Thermal Heatmap Source Localization Benchmark
## 📝 Summary
**ThermBench** is a physics-inspired **synthetic dataset** designed to evaluate algorithms that infer **hidden thermal sources** from an observed **heat diffusion map**.
Each data sample contains:
- an **observed heatmap** (matrix of values),
- and the **ground-truth sources**: `(row, col, intensity)`.
Diffusion follows inverse Manhattan distance:
\[
H(i,j) \;=\; \sum_{s=1}^{K} \frac{I_s}{d(i,j,s)+1}
\]
where \(d\) is the Manhattan distance to source \(s\).
---
## 📊 Dataset Structure
- **level**: Difficulty tier (`very_easy`, `easy`, `medium`, `hard`, `extreme`)
- **input_text**: Heatmap formatted as:
```
N M
K
<N rows of values>
```
- **output_text**: True source positions and intensities in format:
```
row col intensity
```
### Example
```json
{
"level": "easy",
"input_text": "5 5\n2\n10 8 6 5 4\n8 10 7 6 5\n6 7 10 7 6\n5 6 7 10 8\n4 5 6 8 10",
"output_text": "1 1 10.0\n5 5 10.0"
}
```
---
## 🚀 Usage
```python
from datasets import load_dataset
dataset = load_dataset(
"ZoneTwelve/Thermal-Heatmap-Source-Localization",
split="train"
)
print(dataset[0])
```
---
## 🎚 Difficulty Levels
- **very_easy** → 3×3 grid, 1 source
- **easy** → 5×5 grid, 2 sources
- **medium** → 10×10 grid, 3 sources
- **hard** → 20×20 grid, 5 sources
- **extreme** → 30×30 grid, 7 sources
Each level contains 100 samples → **500 total**.
A fuzzy extension of ThermBench introduces noise, intensity jitter, and rounding differences to simulate real‑world sensor readings.
---
## 🔧 Intended Applications
- Benchmarking **inverse problem solvers**
- Robustness studies for optimization/AI
- Educational resource for algorithm development
---
## 📜 License
Apache License 2.0 © ZoneTwelve
|