SadilKhan/MARVEL_FX3D

MARVEL-FX3D

Sankalp Sinha👨‍💻 · Mohammad Sadil Khan👨‍💻 · Muhammad Usama · Shino Sam · Didier Stricker · Sk Aziz Ali · Muhammad Zeshan Afzal

👨‍💻 Equally contributing first authors

---

This repo contains weights for fine-tuned Stable Diffusion 3.5 Large on MARVEL-40M+ dataset. Given a text prompt, the model generates an image suitable for a pretrained image-to-3D model such as Sam3D, Trellis, or Stable Fast 3D.

Inference

# Generate Image from text prompts

import torch
from diffusers import StableDiffusion3Pipeline

model_id = "stabilityai/stable-diffusion-3.5-large"
lora_path = "SadilKhan/MARVEL_FX3D"  # or local path

pipe = StableDiffusion3Pipeline.from_pretrained(
    model_id,
    torch_dtype=torch.float16,
    device_map="auto"
)

# Load LoRA weights
pipe.load_lora_weights(lora_path)

pipe.to("cuda")

prompt = "An old, moss-covered wishing well. Rough stones, aged wood, rusty chains, mushrooms, fallen leaves, and twigs create an enchanting, ancient, and rustic atmosphere."

image = pipe(
    prompt=prompt,
    num_inference_steps=28,
    guidance_scale=7.0,
).images[0]

image.save("output.png")

Citation

If you find MARVEL-FX3D useful, please cite

@inproceedings{sinha2025marvel,
  title     = {MARVEL-40M+: Multi-Level Visual Elaboration for High-Fidelity Text-to-3D Content Creation},
  author    = {Sinha, Sankalp and Khan, Mohammad Sadil and Usama, Muhammad and Sam, Shino and Stricker, Didier and Ali, Sk Aziz and Afzal, Muhammad Zeshan},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={8105--8116},
  year={2025}
}