# Russian Road Signs Dataset

This dataset is designed for training and evaluating traffic sign detection models for autonomous driving. It covers 61 traffic sign classes following the Russian GOST R 52290-2004 standard - warning signs (1.x), priority signs (2.x), prohibitory signs (3.x), mandatory signs (4.x), information signs (5.x), service signs (6.x), and supplementary plates (8.x) - recorded from four vehicle-mounted cameras across 50 driving sessions in urban and suburban environments.

## Dataset Description

- **Source:** Fleet of autonomous vehicles, Russian urban and suburban roads
- **Sensor:** 4x LUCID TRI054S-CC cameras (2880×1860), front/side facing
- **Coverage:** ~54K bounding box annotations, ~6.4K unique images, 61 sign classes
- **Splits:** Train 70% (39,717 rows, 5,031 images), Test 30% (14,564 rows, 1,409 images) - split by driving session to prevent temporal leakage
- **Format:** Parquet (annotations) + JPEG (images) in `train/` and `test/` subdirectories

### Classes (61)

Signs classified according to the Russian GOST R 52290-2004 standard, including:

- **Speed limits and their end signs** — `3_24`, `3_24_s40_white`, `3_24_s50`, `3_24_s60`, `3_24_s70`, `3_24_s70_white`, `3_24_s80`, `3_24_s90`, `3_24_s90_white`, `3_24_s100`, `3_24_s110`, `3_24_s130`, `3_25`, `3_25_white`, `3_25_s30_white`, `3_25_s50`, `3_25_s50_white`, `3_25_s60`, `3_25_s60_white`, `3_25_s70`, `3_24_s10_white`, `3_24_s40`
- **Priority signs** — `2_4` (give way), `3_31` (end of all restrictions)
- **Mandatory direction signs** — `4_1_1` (go straight), `4_2_1`, `4_2_2`, `4_2_3` (obstacle avoidance)
- **Warning signs** — `1_8` (traffic lights), `1_15` (slippery road), `1_16` (rough road), `1_18` (gravel ejection), `1_19` (dangerous shoulder), `1_20_1`, `1_20_2`, `1_20_3` (pedestrian crossing ahead), `1_21` (children), `1_33` (other hazards)
- **Information signs** — `5_1` (motorway), `5_2` (end of motorway), `5_15_2`, `5_15_4`, `5_15_5`, `5_15_6`, `5_15_7` (lane direction)
- **Supplementary plates** — `8_2_3` (action zone end), `8_22_1`, `8_22_2`, `8_22_3` (obstacle type)
- **Miscellaneous** — `table` (speed bump sign), `scoreboard`, `scoreboard_sign` (overhead gantry), `unknown_sign`

### 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 | Traffic sign class label (GOST R 52290-2004) |
| 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 | 39,717 | 5,031 |
| test  | 14,564 | 1,409 |

### Dataset Structure

```
road_signs_dataset/
├── annotations.parquet   # All annotations (54,281 rows)
├── train/
│   ├── camera_1C0FAF5250E2/   # 3,487 images
│   ├── camera_1C0FAF57D6F8/   # 979 images
│   ├── camera_1C0FAF5CA7B6/   # 464 images
│   └── camera_1C0FAF5CC14D/   # 101 images
└── test/
    ├── camera_1C0FAF5250E2/   # 793 images
    ├── camera_1C0FAF57D6F8/   # 149 images
    ├── camera_1C0FAF5CA7B6/   # 364 images
    └── camera_1C0FAF5CC14D/   # 103 images
```

### Usage

```python
import pandas as pd

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