---

Overview

MARVEL-FX3D is the generation component of the MARVEL-40M+ paper, which introduces a dataset of 40M multi-level text annotations for 8.9M+ 3D assets. The dataset and annotation pipeline are described in the main paper.

---

📦 MARVEL-40M+ Dataset

MARVEL-FX3D is trained on MARVEL-40M+, the largest 3D captioning dataset to date.

The multi-stage annotation pipeline combines open-source multi-view VLMs and LLMs with human metadata from source datasets, reducing hallucinations and improving domain-specific accuracy.

🔗 Dataset on Hugging Face: MARVEL-40M+

---

🛠️ Installation

# Clone the repository
git clone https://github.com/SadilKhan/MARVEL-FX3D.git
cd MARVEL-FX3D

# Create a conda environment (recommended)
conda create -n marvel python=3.10
conda activate marvel

# Install dependencies
pip install -r requirements.txt

Requirements: Python ≥ 3.10, PyTorch ≥ 2.0, CUDA ≥ 11.8 (recommended)

---

🚀 Quick Start

# Generate a 3D textured mesh from a text prompt
python generate.py --prompt "A Harley Davidson motorcycle with a black leather seat and dual exhaust pipes"

The output mesh will be saved to output/ by default.

📜 Citation

If you find MARVEL-FX3D or MARVEL-40M+ useful in your research, please cite:

@InProceedings{Sinha_2025_CVPR,
    author    = {Sinha, Sankalp and Khan, Mohammad Sadil and Usama, Muhammad and Sam, Shino and Stricker, Didier and Ali, Sk Aziz and Afzal, Muhammad Zeshan},
    title     = {MARVEL-40M+: Multi-Level Visual Elaboration for High-Fidelity Text-to-3D Content Creation},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2025},
}

---

📄 License

This project is released under the MIT License.

---