---
pretty_name: IALA_B (Daytime, Single Mark JSON→JSON)
license: cc-by-4.0
language:
- en
task_categories:
- text-generation
tags:
- maritime
- navigation
- IALA
- LLM
- VLM
size_categories:
- n<1K
---


# IALA_B: Daytime, Single‑Mark JSON→JSON Dataset (1,500 samples)

**Input** = VLM observation JSON (body/topmark attributes; may include synonyms/unknown).  
**Output** = LLM normalized JSON (mark_type, safe_direction, hazards, consistency, finalized_attributes…).  
Lateral mapping assumes **IALA‑B** (red=starboard, green=port).

- Total samples: **1500** 
- Classes (9): cardinal_east, isolated_danger, lateral_port, lateral_starboard, cardinal_north, safe_water, cardinal_south, special_mark, cardinal_west  
- Input augmentation: **synonyms** for topmark shapes (e.g., *cone-shaped ≈ conical; circular ≈ circle; spheres/balls; apex‑to‑apex ≈ peak to peak*).  
- Outputs are **normalized** to a fixed vocabulary; conflicting/unknown inputs are **downgraded** to `unknown` in `finalized_attributes` and recorded in `consistency.flagged_inconsistent`.

## Schema

### User (input)
```json
{
  "version": "v1",
  "context": { "time_of_day": "day", "region": "IALA-B" },
  "observation": {
    "body_color":    { "value": "<string>", "conf": <float> },
    "topmark_color": { "value": "<string|unknown>", "conf": <float> },
    "topmark_shape": { "value": "<string|unknown>", "conf": <float> }
  }
}
```

### Assistant (output)
```json
{
  "mark_type": "<enum>",
  "safe_direction": "<str>",
  "hazards": "<str>",
  "consistency": {
    "used_signals": ["..."],
    "flagged_inconsistent": ["..."],
    "notes": "<str>"
  },
  "finalized_attributes": {
    "body_color": "(...|unknown)",
    "topmark_color": "(...|unknown)",
    "topmark_shape": "(...|unknown)"
  },
  "confidence": <float>,
  "rules_applied": ["..."]
}
```

## Loading
```python
from datasets import load_dataset
ds = load_dataset("json", data_files={
    "train": "train.jsonl",
    "validation": "validation.jsonl"
})
print(ds)
```