README.md

---
license: apache-2.0
task_categories:
- object-detection
language:
- en
pretty_name: 'SimData-NuScenes: Synthetic Autonomous Driving Dataset'
size_categories:
- 100B<n<1T
---
# Dataset Card for Dataset Name

<!-- Provide a quick summary of the dataset. -->

**SimData-NuScenes** is a large-scale synthetic dataset generated from high-fidelity simulation environments using **aiSim**. 

By leveraging **aiSim's** advanced physics engine and deterministic sensor modeling, we ensure that every frame maintains **high-quality visual fidelity and physical accuracy**. This makes the dataset particularly effective for training and validating perception algorithms where precision is paramount.

The dataset follows the **NuScenes format (v1.0-custom)** and covers diverse environments including highways, complex urban areas, and parking lots across different geographic styles (US, Europe, Japan).


## Dataset Details

### Key Features
- **Format**: Fully compatible with the `nuscenes-devkit`.
- **Scale**: Contains **45 scenes** derived from **15 distinct maps**.
- **Diversity**: Covers Highway, Urban, and Parking scenarios.
- **Volume**: Approximately **18,000+ frames** per sensor (Camera/LiDAR).

### Sensor Layout
    
<div align="center">
  <img src="https://cdn-uploads.huggingface.co/production/uploads/69367df980cb6886b08b3cc9/Uf1-Txyyx2tycRjwP-ZwJ.png" width="80%" />
</div>

### Overview
6 annotated surround-view camera images and BEV ground truth with LiDAR point clouds.
<div align="center">
  <img src="https://cdn-uploads.huggingface.co/production/uploads/69367df980cb6886b08b3cc9/4uB-JKa9_HikgdcZfi31y.jpeg" width="80%" />
</div>

## Dataset Statistics (统计数据)

The dataset metadata is organized as follows:

| Metric | Count |
| :--- | :--- |
| **Total Logs/Scenes** | 45 |
| **Maps** | 15 |
| **Annotated Samples (Keyframes)** | 1,796 |
| **Sample Data (Total Frames)** | 215,472 |
| **Total Annotations** | 64,190 |
| **Ego Poses** | 17,956 |
| **Categories** | 10 |
| **Sensors** | 12 (Cameras, LiDARs, Radar) |

## Object Categories (标注类别)

The dataset includes 3D bounding box annotations for the following **10 classes**:

1. `Car`
2. `Truck`
3. `Bus`
4. `Van`
5. `Trailer`
6. `Pedestrian`
7. `Motorcycle`
8. `Bicycle`
9. `TrafficCone`
10. `Barricade`

## Scenarios & Maps (场景与地图详情)

The dataset is constructed from **15 high-definition maps**, categorized into three main environment types. Each map contains approximately 3 scenarios.

### 🛣️ Highway Environments
| Map Name | Description |
| :--- | :--- |
| **Highway_US-CA_SR85Sunnyvale** | US Highway scenario (SR85), sunny/clear weather. |
| **Highway_US-CA_Construction** | Highway construction zone with barriers and cones. |
| **Highway_HU_Godollo** | European style highway environment. |
| **Highway_US-CA_230Junipero** | Junipero Serra West Walley highway section. |

### 🏙️ Urban Environments
| Map Name | Description |
| :--- | :--- |
| **Urban_US-CA_SanFranciscoCity** | Dense urban downtown environment (SF style). |
| **Urban_US-CA_SF_OuterSunset** | Residential/Suburban area in San Francisco. |
| **Urban_HU_R7BudafokRoundabout** | European urban scene featuring a roundabout. |
| **Urban_Synth_USCity** | Synthetic US city with crowded traffic (`US_CrowdedCity`). |
| **Urban_Synth_USCrossingStreet** | Urban intersection and crossing scenarios. |
| **Urban_Synth_JapanCity** | Japanese style urban environment (LHT - Left Hand Traffic). |
| **Parking_Synth_UrbanSpots_LHT** | Urban street parking scenarios. |

### 🅿️ Parking Environments
| Map Name | Description |
| :--- | :--- |
| **Parking_US-CA_SanJoseMall** | Indoor garage environment. |
| **Parking_US-CA_SanJoseAlamitos** | Outdoor parking lot scenario. |

## How to Use (使用方法)

Since this dataset follows the NuScenes schema, you can use the standard [nuscenes-devkit](https://github.com/nutonomy/nuscenes-devkit) to load and visualize the data.

### Installation
```bash
pip install nuscenes-devkit