# Pillars Dataset

This dataset is designed for detecting roadside vertical structures - pillars, poles, and pylons - which are critical static obstacles for autonomous vehicle navigation. It contains 12K bounding box annotations from two side/rear-facing cameras across 6 driving sessions in urban and suburban environments.

## Dataset Description

- **Source:** Fleet of autonomous vehicles, urban and suburban roads
- **Sensor:** 2x LUCID TRI054S-CC cameras (2880×1860), front/side facing
- **Coverage:** ~12K bounding box annotations, ~1.4K unique images, single class
- **Splits:** Train 70% (11,080 rows, 1,242 images), Test 30% (1,035 rows, 139 images) - split by driving session to prevent temporal leakage
- **Format:** Parquet (annotations) + JPEG (images) in `train/` and `test/` subdirectories

### Classes (1)

| Class | Count | Description |
|-------|-------|-------------|
| pillar | 12,115 | Roadside pillar / pole / pylon |

### Data Fields

| Field | Type | Description |
|-------|------|-------------|
| bbox_msg_id | VARCHAR | UUID linking to the original bounding box message |
| object_id | BIGINT | Unique object tracking ID |
| label | VARCHAR | Object class (`pillar`) |
| bbox_coords | DOUBLE[4] | Bounding box [x, y, width, height] in pixel coordinates |
| timestamp_ns | BIGINT | ROS bag timestamp (nanoseconds) |
| image_path | VARCHAR | Relative path to JPEG in `train/` or `test/` |

### Data Splits

| Split | Rows | Images |
|-------|------|--------|
| train | 11,080 | 1,242 |
| test  | 1,035  | 139   |

### Dataset Structure

```
pillars_dataset/
├── annotations.parquet   # All annotations (12,115 rows)
├── train/
│   ├── camera_1C0FAF5250E2/   # 41 images
│   └── camera_1C0FAF57D6F8/   # 1,201 images
└── test/
    ├── camera_1C0FAF5250E2/   # 40 images
    └── camera_1C0FAF57D6F8/   # 99 images
```

### Usage

```python
import pandas as pd

df = pd.read_parquet("annotations.parquet")
print(f"{len(df)} annotations, {df.image_path.nunique()} unique images")
```