README.md

---
library_name: transformers
license: apache-2.0
language:
- en
base_model:
- Qwen/Qwen2.5-VL-3B-Instruct
tags:
- VLA
- VLM
- LLM
- DriveFusion
- Vision
- MultiModal
- AutonomousDriving
pipeline_tag: robotics
datasets:
- DriveFusion/DriveFusion-Data
---

# DriveFusion-V0.2

<div align="center">
  <img src="drivefusion_logo.png" alt="DriveFusion Logo" width="300"/>
  <h1>DriveFusion-V0.2</h1>
  <p><strong>A Multimodal Autonomous Driving Model: Vision + Language + GPS + Speed.</strong></p>

  [![Model License](https://img.shields.io/badge/License-Apache%202.0-green.svg)](https://opensource.org/licenses/Apache-2.0)
  [![Base Model](https://img.shields.io/badge/Base%20Model-Qwen2.5--VL-blue)](https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct)
  [![Status](https://img.shields.io/badge/Status-Active-success.svg)]()
</div>

---

## 🚀 Model Overview

**DriveFusion-V0.2** is a multimodal model designed for autonomous vehicle applications. Unlike standard Vision-Language models, V0.2 integrates **telemetry data** (GPS and Speed) directly into the transformer architecture to perform dual tasks:
1. **Natural Language Reasoning**: Describing scenes and explaining driving decisions.
2. **Trajectory & Speed Prediction**: Outputting coordinates for future waypoints and target velocity profiles.

Built on the **Qwen2.5-VL** foundation, DriveFusion-V0.2 adds specialized MLP heads to fuse physical context with visual features, enabling a comprehensive "world model" for driving.

## 🔗 GitHub Repository

Find the full implementation, training scripts, and preprocessing logic here:
* **Main Model Code:** [DriveFusion/drivefusion](https://github.com/DriveFusion/drivefusion)
* **Data Collection:** [DriveFusion/data-collection](https://github.com/DriveFusion/carla-data-collection.git)

### Core Features
- **Vision Processing**: Handles images and videos via a 32-layer Vision Transformer.
- **Context Fusion**: Custom `SpeedMLP` and `GPSTargetPointsMLP` integrate vehicle telemetry.
- **Predictive Heads**: Generates **20 trajectory waypoints** and **10 target speed values**.
- **Reasoning**: Full natural language generation for "Chain of Thought" driving explanations.

---

## 🏗 Architecture

DriveFusion-V0.2 extends the Qwen2.5-VL architecture with a modular "Driving Intelligence" layer.

<div align="left">
  <img src="drivefusion_architectures.png" alt="DriveFusion Architecture" width="700"/>
  <p><i>The DriveFusion-V0.2 Architecture: Integrating visual tokens with telemetry-encoded tokens for dual-head output.</i></p>
</div>

### Technical Specifications
- **Text Encoder**: Qwen2.5-VL (36 Transformer layers).
- **Vision Encoder**: 32-layer ViT with configurable patch sizes.
- **Driving MLPs**: 
  - `TrajectoryMLP`: Generates batch × 20 × 2 coordinates.
  - `TargetSpeedMLP`: Generates batch × 10 × 2 velocity values.
- **Context Window**: 128k tokens.

---

## 🚀 Quick Start

Using DriveFusion-V0.2 requires the custom `DriveFusionProcessor` to handle the fusion of text, images, and telemetry.

### Installation
```bash
pip install transformers accelerate qwen-vl-utils torch
```

### Inference Example
```python
import torch
from drivefusion import DriveFusionForConditionalGeneration, DriveFusionProcessor

# Load Model
model_id = "DriveFusion/DriveFusion-V0.2"
model = DriveFusionForConditionalGeneration.from_pretrained(model_id, torch_dtype=torch.float16).to("cuda")
processor = DriveFusionProcessor.from_pretrained(model_id)

# Define Input: Image + Prompt + Telemetry
gps_context = [[40.7128, -74.0060], [40.7130, -74.0058]] # Lat/Lon history
speed_context = [[30.5]] # Current speed in m/s

message = [{
    "role": "user",
    "content": [
        {"type": "image", "image": "highway_scene.jpg"},
        {"type": "text", "text": f"{'-->'*34}Analyze the scene and predict the next trajectory based on our current speed."}
    ]
}]

# Generate
inputs = processor(text=message, images="highway_scene.jpg", gps=gps_context, speed=speed_context, return_tensors="pt").to("cuda")
output = model.generate(**inputs, max_new_tokens=512)

# Results
print("Reasoning:", output["text"])
print("Predicted Trajectory (20 pts):", output["trajectory"])
print("Target Speeds:", output["target_speeds"])
```

---

## 🛠 Intended Use
- **End-to-End Autonomous Driving**: Acting as a primary planner or a redundant safety checker.
- **Explainable AI (XAI)**: Providing human-readable justifications for automated maneuvers.
- **Sim-to-Real Transfer**: Using the model as a sophisticated "expert" driver in simulated environments.

---

## Citation

``` bibtex
@misc{drivefusionqa2026,
  title={DriveFusion-V0.2: A Vision Language Action Model for Autonomous Driving},
  author={Samir, Omar and DriveFusion Team},
  year={2026},
  url={https://huggingface.co/DriveFusion/DriveFusion-V0.2}
}